Download Target Localization in MRI-guided Prostate Biopsy Helen Xu

Target Localization in
MRI-guided Prostate Biopsy
by
Helen Xu
A thesis submitted to the
School of Computing
in conformity with the requirements for
the degree of Doctor of Philosophy
Queen’s University
Kingston, Ontario, Canada
March 2014
Copyright © Helen Xu, 2014
Abstract
Prostate cancer is a worldwide health concern for men. Needle biopsy is the most
definitive form of cancer diagnosis. Target-specific biopsies can be performed under
magnetic resonance imaging (MRI) guidance. However, needle placements are often
inaccurate due to intra-operative prostate motion and the lack of motion compensation techniques. As a result, malignant tumors can be missed, which in turn will lead
to an increased number of repeated biopsies and delaying of treatment. To increase
the needle targeting accuracy, intra-operative prostate motion and deformation need
to be studied so that motion compensation techniques can be developed accordingly.
This thesis intends to make three main contributions:
1. A comprehensive survey of the state-of-art in image-guided prostate needle
placement interventions.
2. Retrospective clinical accuracy validation of a MRI-guided robotic prostate
biopsy system that was used in the U.S. National Cancer Institute for over
6 years. A 3D-3D registration algorithm consists of an initial two-step rigid
alignment followed by a B-spline deformable transform was developed to align
the pre- and post-needle insertion images. A total of 90 biopsies from 24 patients were studied. The mean target displacement, needle placement error, and
clinical biopsy error were 5.2, 2.5, and 4.3 mm, respectively.
i
3. Development of a multi-slice-to-volume registration for intra-operative target
localization. The algorithm aligns the planning volume with three orthogonal
image slices of the prostate acquired immediately before needle insertion. It
consists of a rigid registration followed by a deformable step using only the
prostate region. The algorithm was validated on 14 clinical images sets from
Brigham and Women’s Hospital in Boston, Massachusetts. All registration
errors were well below the radius of a clinically significant tumour (5 mm), and
are considered clinically acceptable.
The results show that there was a substantial amount of biopsy error caused by
prostate motion and deformation during MRI-guided biopsy. This error can be reduced by using quantitative imaging techniques for prostate registration and motion
compensation. In particular, the multi-slice-to-volume registration algorithm demonstrated the feasibility of intra-operative target localization and motion compensation;
which in turn may improve the quality of MRI-guided prostate interventions.
ii
Acknowledgments
I would like to express my sincere gratitude to my advisor Dr. Gabor Fichtinger, who
has brought and guided me through an incredible journey as a graduate student. He
has enriched me both academically and non-academically, and created many learning
opportunities and wonderful memories that I can cherish. I could not have asked for
a better mentor than him.
I would also like to thank my Ph.D. committee Dr. Purang Abolmaesumi and Dr.
Aaron Fenster for their advices and guidance on my research. They have both played
an important role in my graduate studies. I truly appreciate their quick responses
and the delightful conversations we had at conferences.
Next, I would like to say a huge thank you to my colleague Dr. Andras Lasso,
for rescuing me countless times when I got stuck. This thesis would not have been
possible without him. In addition, I would like to thank all of my co-authors who
work or have worked at Johns Hopkins University, U.S. National Institutes of Health,
or Brigham and Women’s Hospital for their time spent on polishing my publications.
I am very grateful for all of the current and past Perk Lab members, who have
shared many joyful and stressful times with me over the course of my studies. They
have created such a positive and vibrant atmosphere for me to work in. I would also
like to thank the School of Computing staff, especially Debby Robertson for taking
iii
care of my administrative and technical issues.
I would like to thank my parents William Xu and Yanhong Ren, and grandparents Chunrong Ren and Xiangmei Zhou for their constant support and pride in my
accomplishments. Finally, I would like to express my gratitude to my husband Dr.
Randall Elzinga for his company, dedication, and encouragement throughout these
years.
iv
Statement of Originality
I hereby declare that the materials and results in this thesis are original and have not
been published elsewhere, except where indicated by specific references.
v
Contents
Abstract
i
Acknowledgments
iii
Statement of Originality
v
Contents
vi
List of Tables
ix
List of Figures
xi
Chapter 1:
Introduction
1.1 Motivation . . . . . .
1.2 Objective . . . . . .
1.3 Contributions . . . .
1.4 Thesis Outline . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Chapter 2:
Background
2.1 Image Guidance Modalities . . . . . . . . . . . .
2.1.1 Ultrasound Guidance . . . . . . . . . . . .
2.1.2 MRI Guidance . . . . . . . . . . . . . . .
2.1.3 CT Guidance . . . . . . . . . . . . . . . .
2.2 Needle Puncture Trajectory . . . . . . . . . . . .
2.2.1 Transrectal Access . . . . . . . . . . . . .
2.2.2 Transperineal Access . . . . . . . . . . . .
2.2.3 Transgluteal and Transischiorectal Access
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
3
3
4
5
.
.
.
.
.
.
.
.
6
6
7
13
14
15
16
18
18
Chapter 3:
Prostate Motion and Deformation Upon Needle Insertion
3.1 Patient-Induced Prostate Motion . . . . . . . . . . . . . . . . . . . .
3.2 Transrectal Procedures . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Transperineal Procedures . . . . . . . . . . . . . . . . . . . . . . . . .
vi
20
21
23
25
3.4
3.5
3.6
Prostate Motion and Deformation Reduction .
3.4.1 Stabilization Needles . . . . . . . . . .
3.4.2 Needle Sizes . . . . . . . . . . . . . . .
3.4.3 Robotic Needle Drivers . . . . . . . . .
3.4.4 Imaging Probe Modification . . . . . .
Intra-procedural Image Registration . . . . . .
3.5.1 Image fusion between TRUS and MRI
3.5.2 Ultrasound Image Registration . . . .
3.5.3 MR Image Registration . . . . . . . . .
Summary . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
26
27
28
29
29
30
30
32
33
35
Chapter 4:
Accuracy Analysis in MRI-guided Robotic Prostate Biopsy 37
4.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.2.1 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.1 Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.2 Image Registration . . . . . . . . . . . . . . . . . . . . . . . . 41
4.3.3 Registration Validation . . . . . . . . . . . . . . . . . . . . . . 44
4.3.4 Biopsy Accuracy Analysis . . . . . . . . . . . . . . . . . . . . 45
4.4 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4.1 Registration Accuracy . . . . . . . . . . . . . . . . . . . . . . 48
4.4.2 Biopsy Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
Chapter 5:
5.1
5.2
5.3
5.4
5.5
Target Localization in MRI-guided Transperineal Prostate
Biopsy using Multi-Slice-to-Volume Registration
55
Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.2.1 Related Works . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3.1 Image Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . 60
5.3.2 Image Registration . . . . . . . . . . . . . . . . . . . . . . . . 62
5.3.3 Registration Validation . . . . . . . . . . . . . . . . . . . . . . 63
Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
5.4.1 Registration Validation . . . . . . . . . . . . . . . . . . . . . . 70
Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Chapter 6:
Conclusion
6.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.1.1 Prostate Motion and Deformation Upon Needle Insertion . . .
vii
79
79
80
6.1.2
6.1.3
6.2
Accuracy Analysis in MRI-guided Robotic Prostate Biopsy . . 81
Target Localization in MRI-guided Transperineal Prostate Biopsy
using Multi-Slice-to-Volume Registration . . . . . . . . . . . . 81
Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Bibliography
84
viii
List of Tables
3.1
Registration parameters of different TRUS and MRI fusion methods .
32
3.2
Deformable registration results of MR prostate images [9] . . . . . . .
35
4.1
Data statistics for accessing biopsy accuracy . . . . . . . . . . . . . .
52
5.1
Registration validation experiments summary . . . . . . . . . . . . .
67
5.2
Summary of execution time statistics for multi-slice-to-volume registration with clinical images . . . . . . . . . . . . . . . . . . . . . . . .
69
5.3
Gradient descent optimizer parameters for the rigid registration step.
69
5.4
L-BFGS-B optimizer parameters for the deformable registration step.
70
5.5
TRE statistics for Experiment 1a: multi-slice-to-volume rigid registration between the planning volume and simulated intra-operative
orthogonal image slices. The simulated images were undistorted (no
deformation was applied)
5.6
. . . . . . . . . . . . . . . . . . . . . . . .
70
TRE statistics for Experiment 1b: multi-slice-to-volume deformable
registration between the planning volume and simulated intra-operative
orthogonal image slices . . . . . . . . . . . . . . . . . . . . . . . . . .
5.7
71
TRE statistics for Experiment 2a: multi-slice-to-volume rigid registration between the planning volume and clinical intra-operative orthogonal image slices . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
ix
73
5.8
TRE statistics for Experiment 2b: multi-slice-to-volume deformable
registration between the planning volume and clinical intra-operative
orthogonal image slices . . . . . . . . . . . . . . . . . . . . . . . . . .
x
73
List of Figures
1.1
a) Anatomy of the male pelvis, including the prostate gland (prostatecancer.ca). b) Zones of the prostate gland (pathologyoutlines.com). .
2
2.1
Anatomical planes (mevislab.de). . . . . . . . . . . . . . . . . . . . .
7
2.2
a) End-fire TRUS probe and an example of the image it produces.
(hitachi-medical-systems.co.uk). b) 3D TRUS image acquisition using
the end-fire probe and its sample image (intechopen.com). . . . . . .
2.3
8
a) Side-fire (bi-plane) TRUS probe and examples of transverse (middle) and sagittal (right) images that it produces. (hitachi-medicalsystems.co.uk). b) 3D TRUS image acquisition using the side-fire probe
and its sample image (intechopen.com). . . . . . . . . . . . . . . . . .
2.4
9
a) Illustration of a 2D TRUS-guided prostate biopsy (cancer.gov) with
an end-fire probe and a sample image (bkmed.com). b) A example of
the 12-core biopsy protocol in the coronal plane of the prostate [4]. .
2.5
9
a) TRUS-guided LDR prostate brachytherapy using a side-fire probe
(duooxfordshire.org.uk). b) Top: Sagittal view of the prostate before
seed implantation. Bottom: Transverse view of the prostate after seed
implantation (bkmed.com). . . . . . . . . . . . . . . . . . . . . . . . .
xi
12
2.6
a) TRUS-guided HDR prostate brachytherapy using a side-fire probe
(intechopen.com). b) Transverse image of the prostate post-needle
insertion [46]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.7
a) Example of a MRI-guided biopsy setup [89]. b) A sample MR image
of the prostate in transverse plane [89]. . . . . . . . . . . . . . . . . .
2.8
2.9
12
13
a) Example of a MRI-guided LDR brachytherapy setup [100]. b) Example of a MRI-guided HDR brachytherapy setup [57]. . . . . . . . .
14
A sample CT scan of the prostate (aboutcancer.com). . . . . . . . . .
15
2.10 a) Transrectal access path (radiographics.rsna.org). b) Transperineal
access path (duooxfordshire.org.uk). c) Transgluteal access path [116]
(transverse MRI with visible needle path). d) Transischiorectal access
path [48] (transverse CT image with visible needle path). . . . . . . .
16
2.11 Patient positions for prostate needle placement procedures. a) Supine
(paramedicine.com). b) Prone (paramedicine.com). c) Lithotomy (hopkinsarthritis.org). d) Left lateral decubitus (www2.aofoundation.org).
3.1
Standard anatomical directions used in human anatomy (thesebonesofmine.wordpress.com). . . . . . . . . . . . . . . . . . . . . . . . . .
3.2
17
22
a) MR image of the prostate without endorectal imaging coil. b) MR
image of the prostate with an air-inflated endorectal imaging coil [64].
c) MR image of the prostate with a special endorectal imaging coil on
the sheath. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3
24
Prostate deformation caused by ultrasound probe positioning during
brachytherapy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
xii
26
3.4
a) Parallel stabilization needle configuration. b) Oblique stabilization
needle configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.5
27
a) Different types of stabilization needles. Top: needle with hook extended. Middle: needle with hook retracted. Bottom: diamond tip
needle [71]. b) Clinical image of the inserted oblique stabilization needles [27]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
28
4.1
The APT-MRI robotic biopsy device system used in NCI [51]. . . . .
39
4.2
Workflow of the three-stage registration algorithm between the preand post-needle insertion volumes using mutual information. . . . . .
4.3
Illustration of the prostate dislocation during needle insertion and the
parameters used in biopsy accuracy analysis. . . . . . . . . . . . . . .
4.4
47
Prostate contour overlays before (left) and after (right) the automatic
registration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.6
47
The decomposition of target displacement into parallel and orthogonal
component. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.5
44
49
Left: Histograms of target displacement (top), needle placement errors (middle), and biopsy errors (bottom) of the 90 biopsies. Right:
Transverse (top), sagittal (middle), and coronal (bottom) view of the
prostate target displacement. ’*’ and ’.’ represent biopsies on the left
and right prostate lobes, respectively. . . . . . . . . . . . . . . . . . .
xiii
51
5.1
Clinical workflow of the image acquisitions during MRI-guided prostate
biopsy. The solid lines indicate the standard biopsy procedure, and the
dashed lines represent the additional orthogonal image slice acquisition
specific to our study. The grey box is the motion compensation registration method we propose to incorporate. . . . . . . . . . . . . . . .
5.2
61
Workflow of the multi-slice-to-volume registration between the preoperative planning volume and intra-operative image slices. . . . . . .
64
5.3
The arrangement of 5 ground truth points chosen on the prostate. . .
65
5.4
Examples of clinical image prostate contour overlay in the transverse
plane. Each of the three images are copies of the same fixed image overlaid with the contours from the moving image a) before registration,
b) after rigid registration, c) after deformable registration. . . . . . .
5.5
68
TREs for Experiment 1a: the 55 multi-slice-to-volume rigid registrations using simulated images. The simulated images were undistorted
(no deformation was applied). A random initial transform of ± 20 mm
translation and ± 10◦ rotation along each axis were set to the registrations. The distance between the initial target and the ground truth
target was defined as the initial misalignment. . . . . . . . . . . . . .
5.6
71
TREs for Experiment 1b: the 11 multi-slice-to-volume deformable registrations using simulated images. The simulated images were warped
± 5 mm based on two sets of manually defined points using Landwrap
landmark deformable registration in 3D Slicer.
5.7
. . . . . . . . . . . .
72
Examples of prostate contour overlay on the simulated fixed transverse
image slice a) before registration, b) after deformable registration. . .
xiv
72
5.8
TREs for Experiment 2a: the 70 multi-slice-to-volume rigid registrations using clinical images. A random initial transform of ± 20 mm
translation and ± 10◦ rotation along each axis were set to the registrations. The distance between the initial target and the ground truth
target was defined as the initial misalignment.
5.9
. . . . . . . . . . . .
74
TREs for Experiment 2b: the 14 multi-slice-to-volume deformable registrations using clinical images. The image volumes were warped ±
5 mm based on two sets of manually defined points using Landwrap
landmark deformable registration in 3D Slicer.
. . . . . . . . . . . .
75
5.10 Experiment 3: Histogram of the target reconstruction distances between multi-slice-to-volume and volume-to-volume registration. . . . .
xv
76
1
Chapter 1
Introduction
The prostate is a walnut sized exocrine gland of the male reproductive system. It
is located in front of the rectum, just below the bladder, and surrounds the urethra
(Figure 1.1). Prostate cancer is a major health concern for men in developed countries.
In 2012, there were an estimated 241,740 new cases and 28,170 deaths of this disease
in the United States (U.S.), making it the most commonly diagnosed cancer and the
second most common cause of cancer death for American men [85].
The two most common screening methods for prostate cancer are prostate-specific
antigen (PSA) test and digital rectal exam (DRE). The existence of a tumour maybe
suspected when there is an increased level of PSA (usually >4 ng/ml [14]) or abnormal
prostate hardness and lumpiness during a DRE. However, PSA has high sensitivity
but low specificity, and results from DRE are often subjective, inconsistent and inconclusive. Therefore, needle biopsy is often recommended to determine the existence
and extent of cancer based on histological analysis. Each year, approximately 1.5
million prostate biopsies are performed in the U.S. and one positive case is found in
every 6 to 8 biopsies. Out of these cancer cases, 1 in 7 patients will die from it [80].
The relatively low positive case and death to biopsy ratio suggests that there is an
2
Figure 1.1: a) Anatomy of the male pelvis, including the prostate gland (prostatecancer.ca). b) Zones of the prostate gland (pathologyoutlines.com).
over biopsy for prostate cancer. This could be a result of the high sensitivity but inconclusive results from PSA testing and DRE [103], thus adding another complexity
dimension to the management of prostate cancer.
There are several ways of treating prostate cancer depending on the stage of cancer and the overall condition of the patient. Early stage cancer can be treated by
prostatectomy, external beam radiation therapy (EBRT), and brachytherapy, while
later stages can be treated with a combination of EBRT and hormone or chemotherapy. Other experimental treatment options include high intensity focused ultrasound
[55], cryotherapy [78] and photodynamic therapy [105].
This thesis focuses on prostate biopsy and other needle placement procedures for
prostate cancer treatment under image guidance.
1.1. MOTIVATION
1.1
3
Motivation
The prostate is a soft tissue organ attached only by connective tissues. Therefore,
needle insertions can be complex when the targeted tissue moves and deforms due
to a combination of mechanical forces exerted by the biopsy device and patient motion. Unfortunately, contemporary standard procedures do not account for these
intra-procedural motions or changes to organ geometry. Inaccurate needle placement
often leads to failure of the procedure and/or complications, such as excessive tissue
damage and bleeding. In order for needle-based interventions to be safe and effective,
it is important to understand the prostate behaviour upon needle insertion and incorporate motion compensation methods into the current clinical protocol to increase
needle placement accuracy.
1.2
Objective
The overall objective of this thesis is to study the intra-operative prostate motion and
deformation upon needle insertion and to develop a motion compensation method for
improving the needle targeting accuracy under magnetic resonance imaging (MRI)
guidance. More specifically, this thesis proposes to:
ˆ Survey the current image-guided prostate needle placement interventions with
emphasis on prostate motion and deformation upon needle insertion
ˆ Retrospectively evaluate the clinical accuracy of a MRI-guided robotic prostate
biopsy system that was used in the U.S. National Cancer Institute (NCI) for
over 6 years
1.3. CONTRIBUTIONS
4
ˆ Develop a registration algorithm for intra-operative target localization and mo-
tion compensation in MRI-guided prostate biopsy
1.3
Contributions
This thesis intends to make three main contributions:
1. A comprehensive survey of the state-of-art in image-guided prostate needle
placement interventions was conducted. This includes reports on intra-operative
prostate motion and deformation as well as motion compensation methods that
have been used clinically. This work is presently in preparation for journal
submission.
2. A three-stage 3D-3D registration algorithm was developed to validate the clinical accuracy of a MRI-guided robotic prostate biopsy system from the U.S. NCI.
Quantitative evaluations of the system and a detailed prostate motion analysis
based on 90 biopsies from 24 patients were reported. This work was published
in Xu et al. IJCARS 2013.
3. A multi-slice-to-volume registration was developed for intra-operative target
localization and motion compensation in MRI-guided prostate biopsy. It was
validated on clinical images from Brigham and Womens Hospital (BWH) in
Boston, Massachusetts. All registration errors and execution times based on 14
MR biopsies images of 10 patients were clinically acceptable. This work was
submitted to IEEE TBME as Xu et al. and is currently under review.
1.4. THESIS OUTLINE
1.4
5
Thesis Outline
The remainder of this thesis is organized as follows:
Chapter 2 reviews the current prostate needle placement procedures for the diagnosis and treatment of prostate cancer.
Chapter 3 reviews studies on intra-operative prostate motion and deformation
upon needle insertion and image registration methods aimed to improve the accuracy
of needle placement.
Chapter 4 presents a 3D-3D registration algorithm developed for accuracy validation of a MRI-guided biopsy system used at the National Cancer Institute(NCI), as
well as the image-based prostate motion analysis.
Chapter 5 introduces a multi-slice-to-volume registration for intra-operative motion compensation and its validation on clinical MR images from Brigham and Women’s
Hospital (BWH).
Finally, Chapter 6 summarizes and concludes the work done for this thesis and
outlines directions for future research.
6
Chapter 2
Background
This chapter introduces the different types of image guidance modalities and needle puncture trajectories used in needle placement procedures for prostate cancer
diagnosis and treatment.
2.1
Image Guidance Modalities
Two-dimensional (2D) transrectal ultrasound (TRUS) is currently the standard imaging modality for guiding prostate needle placement procedures due to its low cost,
ease-of-use, and real-time imaging capabilities. However, magnetic resonance imaging
(MRI) has superior soft tissue image quality, and it has emerged as an alternative
tool for guidance. Although computed tomography (CT) or fluoroscopy are commonly used in radiation treatment planning, it is rarely used as an image guidance
modality for prostate needle placements due to its limited ability to image soft tissues. The following sections explore the pros and cons of each imaging modality for
prostate interventions.
2.1. IMAGE GUIDANCE MODALITIES
7
Figure 2.1: Anatomical planes (mevislab.de).
2.1.1
Ultrasound Guidance
The two most commonly used transrectal ultrasound probes for image guidance are
the end-fire and side-fire probes. The end-fire probe (Figure 2.2) contains a curved
array at the tip, and is typically used for transrectal biopsy. The side-fire probe
(Figure 2.3) has longitudinal transducers, allowing imaging in the sagittal and transverse (Figure 2.1) plane (bi-plane). It is ideal for transperineal procedures, including
brachytherapy.
For 2D TRUS-guided prostate biopsy, the ultrasound probe is equipped with a
needle guide that ensures the puncture trajectory lies in the image plane. It is held
and inserted into the patient’s rectum by the physician to image and to gain access
to the prostate (Figure 2.4 a). Ultrasound images do not show prostate morphology
2.1. IMAGE GUIDANCE MODALITIES
8
Figure 2.2: a) End-fire TRUS probe and an example of the image it produces.
(hitachi-medical-systems.co.uk). b) 3D TRUS image acquisition using
the end-fire probe and its sample image (intechopen.com).
or carcinomas. In addition, artifacts can also be present due to calcifications within
the prostate. Therefore, contemporary TRUS-guided biopsy uses a grid sampling
approach, obtaining around 6 (hence ”sextant biopsy”) to 18 tissue samples from
upper, mid, and lower areas of the left and right sides to obtain a representative
sampling of the gland and determine the degree and extent of cancer [14] (Figure 2.4
b).
Many problems are associated with the aforementioned non-exhaustive systematic
search method for an unknown target. First, the pressure from the transducer probe
while imaging causes dynamic prostate deformation throughout the procedure (Figure
2.4 a), which can lead to inaccurate needle placement. The location of biopsy is also
2.1. IMAGE GUIDANCE MODALITIES
9
Figure 2.3: a) Side-fire (bi-plane) TRUS probe and examples of transverse (middle) and sagittal (right) images that it produces. (hitachi-medicalsystems.co.uk). b) 3D TRUS image acquisition using the side-fire probe
and its sample image (intechopen.com).
Figure 2.4: a) Illustration of a 2D TRUS-guided prostate biopsy (cancer.gov) with
an end-fire probe and a sample image (bkmed.com). b) A example of the
12-core biopsy protocol in the coronal plane of the prostate [4].
2.1. IMAGE GUIDANCE MODALITIES
10
lost after the procedure, making precise re-biopsy of the same region of the prostate
difficult or impossible. TRUS-guided biopsy provides a significant number of falsenegatives and its detection rate is only about 20-40% [94] [52]. Wefer et al. [106]
have also shown that TRUS misses cancer in at least 20% of the cases. A more recent
study by Mozer et al. [61] reported that only 63% of planned targets were reached
under 2D TRUS guidance. Such observations have been seen with no major changes
for about a decade [95] [96]. Cancers have been routinely missed, resulting in a large
number of repeat biopsy cases [107].
To improve the standard 2D TRUS-guided biopsy, Bax et al. [5] and Natarajan
et al. [62] developed a 3D ultrasound-guided (Figure 2.3 b) biopsy tracking and
targeting device. Systematic biopsy (Figure 2.4 b) under 3D guidance was completed
successfully in 180 out of 218 patients. Doppler imaging was also used to determine
possible cancerous regions based on the amount of blood flow within the prostate
prior to biopsy. Cho et al. [11] reported a sensitivity, specificity and accuracy of 80%,
84%, and 82%, respectively, based on 39 patients. Frauscher et al. [28] found Doppler
targeted biopsy detected as many cancers as systematic biopsy with fewer than half
of biopsy cores in 230 patients, while Taverna et al. [93] reported no significant
improvement in detection rate with 300 patients. Takahashi et al. [91] concluded
that Doppler-directed biopsy does not identify cancer with sufficient accuracy to omit
systematic biopsy. Finally, to exploit the features of multiple imaging modalities,
ultrasound and MR imaging have been fused together for biopsy guidance [83] [42]
[13] [62] [97].
2.1. IMAGE GUIDANCE MODALITIES
11
TRUS is also the standard imaging modality for brachytherapy, a commonly practiced localized treatment of prostate cancer. Low-dose-rate (LDR) brachytherapy involves permanent placement of radioactive seeds (Iodine-125) into the prostate via
catheters inserted through the perineum using template guidance (Figure 2.5). Radiation is delivered directly from within the gland while sparing surrounding critical
organs [76]. However, similar to biopsy, using the TRUS probe as an imaging tool
causes distortions to the prostate and affects the accuracy of needle placement. Furthermore, the implanted seeds are difficult to localize in ultrasound images due to
speckles. As a result, the dose distribution is usually calculated based on implanted
catheter position rather than seed position. This is a major drawback since loose
seeds can move within the prostate during or after needle retraction, resulting in
suboptimal dosimetry [72]. Other groups have incorporated fluoroscopy [86], CT [30]
and cone-beam computed tomography (CBCT) [108] into the standard ultrasound
procedure for better seed visualization and allowing for dynamic intra-procedural
dosimetry.
TRUS-guided high-dose-rate (HDR) brachytherapy (Figure 2.6 b) is similar to
LDR, except a higher radiation dose rate is delivered and the radiation sources
(Iridium-192) are removed at the end of the treatment session. In HDR, the catheters
that contain the radioactive seeds remain in place throughout the procedure, whereas
in LDR, catheters are retracted after seed deposition and the seeds may migrate
within the prostate.
2.1. IMAGE GUIDANCE MODALITIES
12
Figure 2.5: a) TRUS-guided LDR prostate brachytherapy using a side-fire probe
(duooxfordshire.org.uk). b) Top: Sagittal view of the prostate before
seed implantation. Bottom: Transverse view of the prostate after seed
implantation (bkmed.com).
Figure 2.6: a) TRUS-guided HDR prostate brachytherapy using a side-fire probe (intechopen.com). b) Transverse image of the prostate post-needle insertion
[46].
2.1. IMAGE GUIDANCE MODALITIES
13
Figure 2.7: a) Example of a MRI-guided biopsy setup [89]. b) A sample MR image
of the prostate in transverse plane [89].
2.1.2
MRI Guidance
MRI provides an alternative approach for image guidance. Its high soft tissue contrast provides clear visualization of the prostate and its anatomical substructures.
This allows suspicious lesions to be identified for target specific biopsies (Figure 2.7).
Several groups have shown that the sensitivity of MRI-guided biopsy exceeds that
of the TRUS-guided system [32] [36] [3]. Estimates of the sensitivity of MRI for the
detection of prostate cancer using T2-weighted sequences and endorectal coils vary
from 60% to 96% [47]. The MRI scanner has strong magnetic fields and confined
physical space; therefore needle insertion may benefit from robotic assistance. Numerous biopsy systems were developed for this purpose [73]. Furthermore, metabolic
information from MR spectroscopy [2], dynamic contrast enhanced (DCE) MRI [88],
and diffusion weighted imaging (DWI)[41] have also emerged for better biopsy target
selection.
2.1. IMAGE GUIDANCE MODALITIES
14
Figure 2.8: a) Example of a MRI-guided LDR brachytherapy setup [100]. b) Example
of a MRI-guided HDR brachytherapy setup [57].
In addition to biopsy, MRI has been used for guiding LDR brachytherapy [15] [37]
[100] (Figure 2.8 a). Due to limited access inside the MR scanner, van Gellekom et al.
[33] developed a single needle method in which a needle is inserted into the prostate
multiple times at different angles instead of the standard multi-parallel approach.
Other groups have also used MRI for treatment planning during HDR brachytherapy
[12] [57] [74] (Figure 2.8 b).
Despite the positive aspects, MRI has not been widely adopted for prostate interventions due to its limited availability and high cost. Nonetheless, it plays an
important part in research and remains as an alternative for patients that are not
suitable for TRUS procedures or have rising PSA levels but repeated negative TRUSguided biopsy results.
2.1.3
CT Guidance
Computed tomography (CT) (Figure 2.9) is not a common imaging modality for guiding needle insertions into the prostate because the organ anatomy is not sufficiently
2.2. NEEDLE PUNCTURE TRAJECTORY
15
Figure 2.9: A sample CT scan of the prostate (aboutcancer.com).
defined in its images [29]. However, it is an option for patients with large prostates
or who are not suitable for transrectal imaging due to prior rectal surgery. Molloy et
al. [60] used CT guidance for brachytherapy. Koutrouvelis et al. [48] [49] performed
both 3D CT-guided biopsy and brachytherapy on over 100 patients.
2.2
Needle Puncture Trajectory
There are four ways to access the prostate with needles: transrectally, transperineally,
transgluteally, and transischiorectally (Figure 2.10). The transrectal approach has
the shortest puncture path and is ideal for biopsy. However, due to infection concerns, treatment procedures that require more extensive needle placements are done
transperineally. The transgluteal and transischiorectal methods are rare and can be
performed on patients with prior rectal surgery. The following sections provide a
more detailed description for each of these methods.
2.2. NEEDLE PUNCTURE TRAJECTORY
16
Figure 2.10: a) Transrectal access path (radiographics.rsna.org). b) Transperineal
access path (duooxfordshire.org.uk). c) Transgluteal access path [116]
(transverse MRI with visible needle path). d) Transischiorectal access
path [48] (transverse CT image with visible needle path).
2.2.1
Transrectal Access
The transrectal approach (Figure 2.10 a) is generally used for biopsy, the definitive
form of diagnosis for prostate cancer. The biopsy needle (usually 18 Gauge) contains
a notch in the distal region and an external sheath that slides over the needle. During
tissue acquisition, the needle enters the targeted region of the prostate and the sheath
closes onto a tissue sample, which is contained in the notch of the needle. Since
2.2. NEEDLE PUNCTURE TRAJECTORY
17
Figure 2.11: Patient positions for prostate needle placement procedures. a) Supine
(paramedicine.com). b) Prone (paramedicine.com). c) Lithotomy (hopkinsarthritis.org). d) Left lateral decubitus (www2.aofoundation.org).
commonly used biopsy needles have cores about 12 to 15 mm long [14], the insertion
depth is a less important factor compared to the transverse needle displacement.
Due to the short puncture path of the transrectal approach, biopsies can be performed with no or local anaesthesia and is normally well tolerated by patients [14].
The left lateral decubitus position (Figure 2.11) is most commonly used for the procedure but some groups have also used prone [89] and supine [20] [21] positions for
transrectal biopsies.
In addition to biopsy, Susil et al. [89] performed fiducial (gold marker) placement
transrectally via needles with the patients in prone position (Figure 2.11). Fiducials
are sometimes needed in external beam radiation therapy (EBRT) for patient position
verification and are often used as reference landmarks for validating image registration
accuracy.
2.2. NEEDLE PUNCTURE TRAJECTORY
2.2.2
18
Transperineal Access
Treatment procedures such as LDR and HDR brachytherapy, which involves a large
number of needle insertions, are commonly performed transperineally (Figure 2.10
b) with the patient in lithotomy position (Figure 2.11). Disinfection of the perineal
region and lumbar or full anaesthesia is required.
Biopsies have also been performed transperineally with the patient in lithotomy
position [16] [35] [8] [66]. The transperineal method is capable of obtaining more tissue
volume from the peripheral zone (PZ), where most of cancer (70%) tends to occur [94].
It has been theoretically demonstrated that of the tissue extracted transperineally,
98.5% can be from the PZ, whereas only 64.9% of transrectally sampled tissue are
from the PZ [59]. However, the transperineal approach requires a much longer needle
path, which increases tissue damage and risks of needle deflection. Based on a study
of 414 patients, Huo et al. [40] reported an average sensitivity and specificity of
48% and 84.1%, respectively, for cancer detection using this approach. There was a
significant decrease in sensitivity for larger prostates.
Furthermore, fiducial placement can be performed transperineally under TRUSguidance similar to the brachytherapy setup. Van den Bosch et al. [100] demonstrated
the feasibility of the procedure under MRI-guidance and robotic assistance with the
patient in lithotomy position.
2.2.3
Transgluteal and Transischiorectal Access
For patients with prior rectal surgery, the transgluteal (Figure 2.10 c) approach can
be used. Zangos et al. [116] carried out biopsies with the patients in prone position.
A large insertion depth is required for this method, but it was mentioned that this
2.2. NEEDLE PUNCTURE TRAJECTORY
19
technique provides superior results for biopsying the apex part of the prostate and it
minimizes injuries to the bladder, bowel loops, and iliac vessels.
The transischiorectal approach (Figure 2.10 d) was used in CT-guided biopsy [48]
and brachytherapy [60] [48] [49] with the patients in prone position.
20
Chapter 3
Prostate Motion and Deformation Upon Needle
Insertion
Over the past decade, methods for improving standard image-guided diagnosis and
treatment of prostate cancer have received significant attention in the research community. However, only a fraction of these have been evaluated in clinical practice.
Numerous systems and algorithms that were developed remained in academia, away
from the operating room. Since the end goal of such projects is to improve the quality of current patient care, clinical compatibility and trial outcomes therefore have
the utmost importance. It would be beneficial to extract the proposed methods that
have made into the operating room from the large pool of prostate imaging literature.
The objective of this chapter is to survey recent literature on the state-of-art in clinical image-guided prostate needle placement procedures with emphasis on prostate
motion and deformation upon needle insertion.
3.1. PATIENT-INDUCED PROSTATE MOTION
3.1
21
Patient-Induced Prostate Motion
Reports from external beam radiation therapy (EBRT) show the amount of intraprocedural prostate motion that already exists without needle insertions or the usage
of transrectal imaging probes. Wallner et al. reported the primary prostate motion in the absence of needle insertion is in the anterior-posterior direction (Figure
3.1). Vargas et al. [102] studied prostate motion in 7 patients during a 4-minute
Cine-MRI. Prostate motion was found to be dependent on time, patient position,
and rectal filling. More motion was observed in the second half of the experiment
compared to the first half. Furthermore, the prostate moved more in prone position
than supine. Van Herk et al. [101] also found larger prostate motion in prone position than supine position based on 11 patients during radiotherapy. Dinkel et al. [18]
reported respiratory-induced mean prostate displacement of 43 patients in cranialcaudal (Figure 3.1) and anterior-posterior direction to be 2.7 ± 1.9 mm and 1.8 ±
1 mm, respectively, using dynamic MRI. In 69% of the subjects, breathing-related
prostate movements were found to be less than 3 mm.
In addition to the motion that occurs naturally, image-guided needle insertions
further deforms the prostate. The accuracy of needle placement procedures is affected
by a combination of the following:
ˆ Pressure induced by the transrectal imaging probe [44] [4]
ˆ Tissue motion and deformation during needle insertion and tissue acquisition
[112] [82]
ˆ Needle bending and deflection during insertion [8]
ˆ Time delay between imaging and needle insertion [89] [102]
3.1. PATIENT-INDUCED PROSTATE MOTION
22
Figure 3.1: Standard anatomical directions used in human anatomy (thesebonesofmine.wordpress.com).
ˆ Reactionary patient motion due to discomfort (leg motion, clenching of pelvic
floor muscles, pelvis movement) [112] [63]
ˆ Unintentional patient motion (breathing, rectum activity, bladder filling) [63]
The extent of these motions also depends on patient position and varies from
patient to patient. Observations have shown that the maximum prostate deformation
upon needle insertion occurs during capsular penetration stage of the prostate [58].
Considering these factors, needle placements still need to be sufficiently accurate
to hit the targeted tissue in order for the procedures to be effective. The clinical
significance of a tumor is dependent on many factors such as its size, location, and
Gleason score. For the purpose of this thesis, we define a clinically significant tumor
to be one with a volume of 0.5 cm3 or greater [69], which corresponds to a spherical
tumour with a radius of approximately 5 mm. Although tumour shapes can be irregular, this sets an upper limit for needle targeting accuracy. To improve the accuracy
3.2. TRANSRECTAL PROCEDURES
23
of needle placements, prostate behaviour upon needle insertions need to be studied.
The rest of this chapter discusses prostate motion and deformation during transrectal and transperineal needle placement procedures as well as methods attempted to
reduce these affects.
3.2
Transrectal Procedures
During transrectal procedures, the prostate shape and position are affected by the
presence of imaging probe in the rectum and forces from the needle insertion. In addition, for TRUS-guided procedures, the probe exerts varying pressure on the rectum
and prostate while imaging, whereas in MRI-guided procedures, there is time delay
between imaging and needle insertion. The following reviews prostate dislocation and
deformation under both TRUS and MRI guidance.
Majority of the prostate motion and deformation in TRUS-guided biopsy are
caused by the imaging probe (Figure 2.10 a). The prostate moves and deforms as a
result of the probe placement inside the rectum and the varying pressure it exerts on
the prostate while imaging. It was found that the prostate dislocation is predominately along the transducer probe axis and the majority of the deformation occurs in
the peripheral zone (PZ) (Figure 1.1 b) [44].
For motion due to needle insertion in TRUS-guided biopsy, De Silva et al. [82]
performed non-rigid registration on data from 9 patients and observed that the mean
dislocation as a consequence of needle insertion and tissue acquisition alone is only
0.4 mm. The majority of this occurred lateral to the needle, away from the center of
the prostate, and at the needle’s piercing point at the prostate boundary.
The presence of endorectal imaging coil used in MRI-guided transrectal biopsies
3.2. TRANSRECTAL PROCEDURES
24
Figure 3.2: a) MR image of the prostate without endorectal imaging coil. b) MR
image of the prostate with an air-inflated endorectal imaging coil [64]. c)
MR image of the prostate with a special endorectal imaging coil on the
sheath.
also deforms the prostate (Figure 3.2). However, the coil should ideally be stable
after its initial placement and therefore decoupling prostate motions from itself [51].
An exception to this is the device developed by Elhawary et al. [19], where the probe
containing the coil is able to move transrectally. In addition, Beyersdorf et al. [6]
performed biopsy with an in-vivo device that only used a surface body coil instead
of an endorectal probe. Apart from imaging probe issues, there is a long time gap
between image acquisition and needle insertion [89], during which prostate motion
can occur.
To quantify prostate motion during biopsy under MRI guidance, Xu et al. [109]
performed a validation study on an MRI-guided robotic biopsy system developed by
Krieger et al. [51]. The paper uses an intensity-based two-step rigid registration
followed by a deformable refinement between the pre-and post-needle insertion image
volumes to capture prostate motion during the procedure. It was observed that the
mean tissue displacement of 90 biopsies was 5.2 mm. The results show that even
3.3. TRANSPERINEAL PROCEDURES
25
though robotic needle placement was accurate, the dislocation of the intended target
can still lead to a mean biopsy error of 4.3 mm. To determine whether this movement
is related to the needle insertion direction, the displacement was decomposed into two
components: one parallel and one orthogonal to the needle vector. It was found that
the two were not significantly different from each other.
3.3
Transperineal Procedures
Transperineal prostate procedures are generally performed with the patient under
general or lumbar anaesthesia, therefore patient motion in this case is reduced. However, prostate deformation is still present due to needle insertions and tissue acquisitions. In addition, this approach still requires the presence of an imaging probe for
prostate visualization, which deforms the anatomy and displaces the prostate. The
following discusses effects of transperineal needle insertions on prostate motion and
deformation.
Regarding the effects of needle insertion during TRUS-guided LDR brachytherapy, Stone et al. [87] found significant gland deformation and displacement in 19
patients. The medium movement in cranial-caudal, lateral, and anterior-posterior direction was 15 mm, 1.9 mm and 2.6 mm, respectively. Wallner et al. [104] reported up
to 10 mm of lateral and cephalad motion for needle insertion during ultrasound and
fluoroscopy-guided LDR brachytherapy. Furthermore, anterior displacement occurs
when posteriorly placed needles strike the posterior surface tangentially. Lagerburg
et al. [53] have shown significant prostate rotation upon needle insertion in the coronal plane using 3D ultrasound. The maximum rotation was 13.8° with a significant
correlation between the place of insertion and rotation. In the sagittal plane, the
3.4. PROSTATE MOTION AND DEFORMATION REDUCTION
26
Figure 3.3: Prostate deformation caused by ultrasound probe positioning during
brachytherapy.
maximum rotation was 10.2°. Furthermore, only rotation in the coronal plane could
be predicted and reduced by the use of locking needles.
The insertion depth of the TRUS probe need to be adjusted to image the prostate
from its base to apex during brachytherapy, and therefore deforming the prostate
(Figure 3.3). Different types of MRI endorectal imaging coil can also induce various
deformations of the prostate. Kim et al. [45] reported that inflatable coils may cause
larger distortions (4.1 ± 3 mm) in the anterior-posterior direction than rigid coils (1.2
± 2.2 mm).
Intra-operative prostate motion during transperineal MRI-guided prostate biopsy
was quantified by Fedorov et al.. Based on 538 images from 40 patients, it was found
that the mean prostate centroid motion was 8.7 mm (range: 0.2-34.7 mm).
3.4
Prostate Motion and Deformation Reduction
There are several methods used in clinical practice to reduce prostate motion. However, none of them is capable of completely eliminating motion or deformation. This
section lists these motion reduction techniques.
3.4. PROSTATE MOTION AND DEFORMATION REDUCTION
27
Figure 3.4: a) Parallel stabilization needle configuration. b) Oblique stabilization
needle configuration.
3.4.1
Stabilization Needles
Bilateral stabilization (Figure 3.4 and 3.5) needles are sometimes used in brachytherapy to reduce prostate motion, especially in the transverse plane, in order to have
more precise seed placement and dose distribution. The stabilizing needles are inserted into the outer left and right side of the prostate prior to therapy and remain
throughout the procedure to provide rigidity and support to the organ. However,
Taschereau et al. [92] found no effect of parallel stabilization needles in reducing seed
misplacement by comparing the results from 10 experimental and 20 control patients.
Feygelman et al. [26] discovered an improved treatment quality using three instead
of two stabilizing needles to secure the prostate. Dattoli et al. [17] inserted the
stabilizing needles obliquely into the prostate of a patient and the maximum lateral
displacement decreased from about 10 mm to 0.2 mm. The motion in cranial-caudal
direction also became negligible. In a more recent study, Podder et al. [71] observed
25 brachytherapy patients and found about 24% and 15% reductions in prostate
movement when using two 17G and 18G stabilization needles, respectively. Further
phantom study by the same group revealed that the oblique configuration of needles
3.4. PROSTATE MOTION AND DEFORMATION REDUCTION
28
Figure 3.5: a) Different types of stabilization needles. Top: needle with hook extended. Middle: needle with hook retracted. Bottom: diamond tip needle
[71]. b) Clinical image of the inserted oblique stabilization needles [27].
was more effective than the parallel configuration, and hooked needles provided more
stabilization compared to regular brachytherapy needles [71].
3.4.2
Needle Sizes
Regarding needle sizes, Podder et al. [70] reported a lower insertion force and velocity
for 18G needles compared to 17G based on data from 20 brachytherapy patients.
Therefore, 18G needles cause less prostate deformation during procedure. However,
17G needles are used for patients with harder prostate due to prior therapies. Some
transverse forces were observed on the needles, which were due to the combined effects
of lateral forces exerted by the movement of internal organs, heterogeneity of tissues,
3.4. PROSTATE MOTION AND DEFORMATION REDUCTION
29
presence of template, pre-bending of the needle, and lateral movement of the surgeon’s
hand.
3.4.3
Robotic Needle Drivers
Lagerburg et al. [54] developed a new tapping needle insertion method to reduce tissue
motion. Prostate displacement due to needle insertion was measured using ultrasound
images on video record in 30 patients. A total of 32 needles were manually pushed
into the prostate and 29 were tapped with a prototype robotic system. The mean
prostate motion in cranial-caudal direction was significantly less when the needle was
pushed (5.6 mm) than when the needle was tapped (0.9 mm) into the prostate. Other
robotic needle drivers with rotation, oscillation, translation, tapping, or a combination
of these techniques have also been proposed but not clinically tested.
3.4.4
Imaging Probe Modification
To address the prostate deformation problem caused by pressure from transrectal
imaging probes, devices were built to decouple the probe motion from that of the
prostate. Bax et al. [5] developed a passive mechanical arm to provide 3D ultrasound
image reconstruction by rotational scanning without distorting the prostate. Clinical
trial results from 3 patients have shown in-plane prostate displacements of less than
1 mm during the biopsy procedure. In MRI-guided procedures, Krieger et al. [51]
designed a protective sheath to physically separate robot end-effector motion from
the prostate.
3.5. INTRA-PROCEDURAL IMAGE REGISTRATION
3.5
30
Intra-procedural Image Registration
The above findings suggest that pre-treatment images alone may not be sufficiently reliable for accurate needle placement. It is useful to integrate pre- and intra-procedural
data to improve the overall targeting accuracy. Although many groups have proposed
methods to address the problem, we mainly focus on the ones that have been used in
clinical practice. A few other methods that are part of ongoing clinical trials which
specifically concentrate on dealing with intra-procedural prostate motion during needle placement are briefly mentioned.
3.5.1
Image fusion between TRUS and MRI
TRUS has real-time imaging capabilities but poor image quality while MRI has superior soft tissue contrast but slow image acquisition speed. Therefore by combining
the two, advantages from both imaging modalities can be exploited.
For transrectal biopsy guidance, Xu et al. [114] presented a system that fuses
real-time 2D TRUS images with pre-acquired 3D MRI. Singh et al. [83], Pinto et
al. [68], and Kadoury et al. [42] all demonstrated the feasibility of this system
and showed an increased positive biopsy rate on a cohort of 101 patients. The system
uses electromagnetic tracking and intra-procedural image registration to superimpose
the MRI data onto the US images. It was shown that the overlap of prostate in
MRI and US images was 71 ± 18% without motion compensation and 90 ± 7%
after motion compensation. Similarly, Cool et al. [13] fused MRI and TRUS by a
surface-based non-linear registration incorporating thin-plate splines to provide realtime overlays of suspicious MR lesions onto 3D TRUS. Results of the initial clinical
study involving 25 patients demonstrated a significantly higher rate of positive biopsy
3.5. INTRA-PROCEDURAL IMAGE REGISTRATION
31
cores and a higher Gleason score cancer grading compared to the standard 12-core
sextant biopsy. Natarajan et al. [62] incorporated pre-procedural MRI with DCE and
DWI for biopsy target selection prior to the fusion with real-time ultrasound. The
fusion algorithm consists of segmentation, rigid alignment via anatomical landmarks,
surface registration, and elastic interpolation. A 33% positive biopsy rate was found
in 47 patients compared to only 7% in systematic biopsy.
Kaplan et al. [43] performed targeted transperineal biopsy in 2 patients by fusing
intra-procedural TRUS with pre-procedural MRI. A 3D affine transformation was
performed on 6 manually chosen corresponding fiducials on the ultrasound and MR
images. The computation took less than 1 second to complete and 7 out of 8 biopsies
were found to be positive.
In brachytherapy, Reynier et al. [75] fused real-time ultrasound and pre-procedural
MR data of 11 patients during intra-procedural dosimetry planning. The fusion was
accomplished by a two-stage surface-based registration, with an initial rigid step followed by elastic registration. Patient trials have shown that this method improves
TRUS image segmentation, especially in the apex and base of the prostate, which is
important in treatment planning.
Table 3.1 summaries the registration parameters of all ultrasound to MRI fusion
methods mentioned in this section.
The following methods concentrate on dealing with intra-procedural prostate motion during needle placement. They have not yet been integrated into clinical practice,
but they are all part of ongoing clinical research projects and have been validated on
retrospective patient images with promising results.
3.5. INTRA-PROCEDURAL IMAGE REGISTRATION
32
Table 3.1: Registration parameters of different TRUS and MRI fusion methods
Author
Xu
[114]
Cool
[13]
Application
transrectal
biopsy
transrectal
biopsy
Similarity
Metric
SSD
MI
Optimizer
Transform
Gaussian
Newton
na
rigid
rigid and
thin-plate
spline
rigid and
elastic
affine
Natarajan
transrectal
na
na
[62]
biopsy
Kaplan
transperineal
SSD
simulated
[43]
biopsy
annealing
Reynier
brachytherapy
SSD
Levenberg
rigid and
[75]
variation
Marquardt
elastic
Abbreviations: SSD = sum of square distance; MI = mutual information; na = not
available.
3.5.2
Ultrasound Image Registration
To improve the targeting accuracy in standard 2D ultrasound guidance, De Silva et
al. [81] proposed an intensity-based (normalized cross correlation) prostate tracking
technique that rigidly registers real-time 2D images with a pre-acquired 3D reference
volume. The method was validated using 33 manually identified intrinsic fiducials
in 8 subjects and the target registration error was found to be 1.9 mm. Similarly,
Baumann et al. [4] used a 2D-3D registration approach capable of deformation estimation. It solves the patient motion problem with an a priori model of rectal probe
kinematics, and prostate deformations are estimated with rigid and elastic registrations. Correlation coefficient and correlation ratio were used as the similarity metrics
for rigid and deformation registration, respectively. The accuracy was evaluated to
be 0.8 ± 0.5 mm using manually segmented fiducials from 40 patients.
Registration accuracy is affected by the type of method used. Karnik et al. [44]
3.5. INTRA-PROCEDURAL IMAGE REGISTRATION
33
evaluated prostate surface- and image-based 3D-to-3D TRUS registration by measuring the target registration error (TRE) of manually marked, corresponding, intrinsic
fiducials in the whole gland and PZ. The study found that image-based registration
method yields significantly lower error than the surface-based approach. Since the
central region of the PZ moves upward and the lateral lobes bulge downward in response to probe pressure, the PZ had an increased error anisotropy compared to the
whole gland.
A comparison between rigid and elastic ultrasound image registration was carried
out by Baumann et al. [4] to evaluate the clinical relevance of deformation estimation.
It was found that elastic registration improved TRE of rigid registration about at least
one third. However, the gland is barely deformed in majority of the volumes that
were analyzed. Only about 20% to 30% of volumes have strong deformations that
can be observed and reduced with elastic registration. There was also a computation
time increase of three to four folds for elastic registration in order to reduce the risk
of a missing target by only 10%.
3.5.3
MR Image Registration
To address the prostate motion problem that occurs due to time delay between
MRI acquisition and actual needle biopsy, Tadayyon et al. [90] proposed an intraprocedural multi-slice-to-volume registration to compensate for the change. Three orthogonal intra-procedural slices are acquired in the approximate center of the prostate
and are registered to a high-resolution target planning volume using mutual information (MI).
Fei et al. [24] investigated the difference between rigid and deformable registration
3.5. INTRA-PROCEDURAL IMAGE REGISTRATION
34
in MR images. The rigid registration uses MI and correlation coefficient as similarity
measures at different resolutions. The non-rigid registration is based on independent
optimization of many interactively placed control points using MI and thin plate
spline transformation. More than 100 registration experiments with 17 MR volume
pairs were performed. Both visual and numerical evaluation confirmed that non-rigid
registration consistently worked better than rigid body. In addition, results showed
that <180 strategically placed points were sufficient to capture the features of prostate
deformation. Tadayyon et al. [90] also concluded that non-rigid registrations of MRI
yield more accurate results, but rigid tracking was sufficient enough in capturing
majority of the motion during intervention.
Another study by Brock et al. [9] accessed the accuracy, reproducibility and
computational performance of different deformable image registration algorithms on
MR prostate images. Based on results provided by three different institutions using
the same data set, it was found that the linear elastic finite element analysis method
achieved the best results but took the longest time to perform. The results are
summarized in Table 3.2.
It is worth mentioning that there are extensive research focusing on the registrations between clinical data and prostate models for motion or deformation prediction.
These include statistical shape [38], finite element [7], and statistical motion [39] models. Some groups also focus on the registration of pre- and intra-procedural images
[65], while others fuse clinical data with histopathology images to identify cancerous
regions [56]. However, these methods are not necessarily specific to dealing with intraprocedural prostate motion during needle placement and none have been implemented
clinically.
3.6. SUMMARY
35
Table 3.2: Deformable registration results of MR prostate images [9]
Method
Thin plate
Linear
spline
elastic FEA
Time (min)
0.12-5
1.2-22
Similarity
surface
contour
Metric
matching
matching
AVG MAG
7.4 mm
4.3 mm
2.3 mm
ABS AVG LR
6.2 mm
3.1 mm
2.0 mm
ABS AVG AP
0.5 mm
3.7 mm
2.0 mm
ABS AVG SI
1.0 mm
3.2 mm
0.4 mm
Abbreviations: FEA = finite element analysis; SSD = sum of square distance; AVG
= average; MAG = magnitude; ABS = absolute; LR = left-right; AP =
anterior-posterior; SI = superior-inferior.
3.6
Demons: active force
and multi-resolution
0.75-2.3
SSD
Summary
Overall, findings of image-guided needle placement procedures for prostate cancer
have shown that there is substantial prostate dislocation and deformation intraoperatively. However, current standard ultrasound-guided procedures do not account
for these changes, which result in inaccurate needle placements. Although some studies have been done to improve the basic clinical protocols, ultrasound is still limited in
its image quality. In addition, the varying probe pressure exerted on the prostate while
imaging makes it difficult to quantify, model, and predict intra-procedural prostate
position.
Recent research have moved onto MRI for solutions to these problems. Even
though MRI have been used with and without ultrasound in guiding needles into the
prostate, it is still at its early stages of research since many questions remain unsolved.
Very few studies have quantified prostate motion and deformation for MRI-guided
interventions. This information can provide a basis for statistical analysis or modelling
and then be used as a reference to predict intra-procedural prostate location upon
3.6. SUMMARY
36
needle insertion. Once the prostate movement is predictable, motion compensation
techniques such as tracking and registration between pre and intra-procedural images
can be developed and integrated into the operating room to improve needle targeting
accuracy. In general, algorithms that works well are usually more computationally
expensive. Therefore, knowing how much accuracy can be sacrificed for computation
speed in real time procedures is important.
Another challenge is to find validation methods to test the accuracy and robustness of these intra-procedural registration algorithms. The prostate does not contain
anatomical landmarks such as bones to be used for registration evaluation. As a
result, contours are often used as an assessment. Contouring of the prostate can be
subjective, inconsistent, and difficult especially at the apex and base of the gland.
The question of how accurate the contours need to be for validation or even registration purposes is still open to discussion. In conclusion, a lot of work has been done in
prostate cancer research over the past decade, yet there is still a lot more that need
to be done in the next decade.
37
Chapter 4
Accuracy Analysis in MRI-guided Robotic
Prostate Biopsy
The material in this chapter has been published as Xu et al. in the International
Journal of Computer Assisted Radiology and Surgery (IJCARS) [109]. Preliminary
studies of this work were also published in Medical Image Computing and Computer Assisted Intervention (MICCAI) [112], Third Hamlyn Symposium on Medical
Robotics [111], Medical Physics [113], and SPIE Medical Imaging 2010 [110], where
I was the first author.
4.1
Overview
Purpose: To assess retrospectively the clinical accuracy of an MRI-guided robotic
prostate biopsy system that has been used in the U.S. National Cancer Institute for
over six years.
Methods: Series of 2D transverse volumetric MR image slices of the prostate both
pre (high-resolution T2-weighted) and post (low-resolution) needle insertions were
used to evaluate biopsy accuracy. A three-stage registration algorithm consisting
4.2. INTRODUCTION
38
of an initial two-step rigid registration followed by a B-spline deformable alignment
was developed to capture prostate motion during biopsy. The target displacement
(distance between planned and actual biopsy target), needle placement error (distance
from planned biopsy target to needle trajectory), and biopsy error (distance from
actual biopsy target to needle trajectory) were calculated as accuracy assessment.
Results: A total of 90 biopsies from 24 patients were studied. The registrations
were validated by checking prostate contour alignment using image overlay, and the
results were accurate to within 2 mm. The mean target displacement, needle placement error, and clinical biopsy error were 5.2 mm, 2.5 mm, and 4.3 mm, respectively.
Conclusion: The biopsy error reported suggests that quantitative imaging techniques for prostate registration and motion compensation may improve prostate
biopsy targeting accuracy.
4.2
Introduction
The superior soft tissue imaging quality of MRI provides clear visualization of the
prostate along with its substructures, such as the peripheral zone, where most of
cancer occurs [94]. It allows suspicious lesions to be identified so that target specific
biopsies can be performed at these sites. MRI-guided biopsy can be performed on patients who have had prior negative outcomes from TRUS-guided biopsy but persistent
cancer symptoms such as high levels of prostate specific antigen (PSA). Patients with
previously diagnosed cancer can also use MRI guidance for re-biopsying the tumor
site to detect if there are any cancer reoccurrence. Due to the limited physical space
in conventional closed-bore scanners and the length of the procedure, robotic assistance is often required. Numerous MRI-compatible biopsy systems were developed
4.2. INTRODUCTION
39
Figure 4.1: The APT-MRI robotic biopsy device system used in NCI [51].
for this purpose [31].
The Access to Prostate Tissue under MRI (APT-MRI) system has been used at
the U.S. National Cancer Institute (NCI) for over six years [51] [50] (Figure 4.1). The
robot is fixated to the patient table and the end that contains the imaging probe
with a built-in needle guide is placed inside the patient transrectally. The device is
then calibrated to scanner coordinate system, and diagnostic scans are taken. Next,
it is remotely controlled to set the desired needle position and angle for a specific
biopsy target. The needle is then advanced into the prostate transrectally through
the needle guide to acquire tissue samples for histological analysis.
The prostate movement upon needle insertion can be extremely complex since
it can deform and dislocate independently from surrounding structures. In addition,
patient movement due to discomfort can further complicate the problem. The current
system does not take into consideration of these factors, yet the biopsies still need to
be sufficiently accurate to hit the intended target in order not to miss the suspected
cancerous tissue. This paper reports a retrospective quantitative evaluation of the
biopsy accuracy for the APT-MRI robotic biopsy system. In addition, a detailed
prostate motion analysis during biopsy is also provided.
4.2. INTRODUCTION
4.2.1
40
Related Works
Prostate motion and deformation upon needle insertion have only been studied by a
few groups using MR images. Some common approaches include tracking a number
of manually identified anatomical landmarks [43] or using surface contours to align
the prostate [37] [75]. The accuracy of these feature-based methods depends heavily
on the user segmentation, which can be inconsistent especially at the apex and base
of the prostate gland. Some groups applied biomechanical models to study the organ
geometry and boundary constraints [10] [1] [58], while others chose image-based methods such as rigid or deformable registration using mutual information and correlation
coefficient [65] [25]. Biomechanical models also require segmentation and knowledge
of material properties, which can be difficult and time consuming. Thus, image-based
rigid and deformable registration would be the most suitable for our case. However,
due to huge variability in the image quality from the large dataset provided by the
NCI, none of the existing method mentioned is capable of accurately determining
the transformation between our image pairs with short computation time and little
manual interference. We developed an algorithm designed to capture the majority of
prostate motion during APT-MRI-guided transrectal biopsy for most of our patient
data.
A preliminary study was previously done by our group [112]. This version presents
results from a larger data set and includes several major improvements to the original
registration framework. These include: image pre-processing, deformable registration
and accuracy validation on all dataset instead of a randomly selected small subset,
validation using ground truth, more in depth statistical analysis, and major decrease
in the amount of manual work.
4.3. METHODS
4.3
41
Methods
4.3.1
Data Acquisition
The MR images were collected from the U.S. NCI over a period of six years. Although
there were variations in the clinical protocol, the following steps were common to
all trials. First, a series of 2D high-resolution T2 transverse volumetric image slices
covering the whole prostate were acquired with the patient in prone or supine position
inside the MRI scanner. From this pre-needle insertion volume, the clinicians select
the biopsy target locations in right-anterior-superior (RAS) coordinates, where the
origin is approximately the center of the prostate. Once the biopsy target locations
are chosen, the APT-MRI device was used to place the biopsy needle transrectally into
the prostate to acquire tissue samples. While the needle is still in place, another set
of 2D transverse volumetric image slices were obtained to confirm needle placement.
There was an at least 10-minute gap between the diagnostic targeting image acquisition and the biopsy needle confirmation image, during which prostate and patient
movement may have occurred. Therefore, to obtain the actual biopsy target location, image registration between the pre and post needle insertion volumes need to
be performed to account for rigid motion and deformation during the procedure. The
resulting transformation from the registration can then be applied to the planned
biopsy target to locate its coordinates in the post needle insertion volume.
4.3.2
Image Registration
Developing a registration algorithm to capture the prostate motion and deformation
for the majority of the images in the dataset was a difficult task. The images were
collected over the past 6 years from different clinical trials, using different imaging
4.3. METHODS
42
protocols on different scanners, by different clinicians, with several versions of the
APT-MRI device. Therefore, there are large variations among the images, including
resolution, artifacts etc. In addition, the complex prostate movement and deformation due to needle insertion along with patient motion during the procedure further
complicate the task. The extent of these motions and deformations also varies from
patient to patient. Nonetheless, a computation method that is suitable for most of
the cases is needed for retrospective biopsy accuracy analysis. The rest of this section
describes our implementation details.
The data used for registration and biopsy accuracy evaluation are the sets of 2D
transverse volumetric image slices of the prostate pre and post needle insertion. The
MR images were first pre-processed to decrease intensity non-uniformity in homogeneous tissue regions using N4ITK (Nicks N3 Insight Toolkit) implementation for
MRI bias field correction [99]. This method does not require expert supervision, user
interaction or training, and only has a few user-defined parameters. The two most important parameters are bias full width at half maximum (BWHM) and noise. BWHM
defines the Gaussian that estimates the bias field, and noise specifies the Wiener filter
used for field estimation. By experimentation, it was found that BWHM at 0.5 and
noise at 0.01 worked the best for most of our clinical images. The remaining few
images had better results with a noise of 0.1. Other parameters had a much smaller
influence on the bias correction results.
After pre-processing, a three-stage volume-to-volume registration procedure was
developed using ITK [115] to determine the transformation between the pre and
post needle insertion volumes. This captures the prostate movement, including both
dislocation and deformation during biopsy (Figure 4.2). The procedure starts with
4.3. METHODS
43
a simple rigid registration of the entire image volume to compensate for prostate
motion in coherence with the biopsy device and patient. Next, another rigid step
was performed using only the prostate as the region of interest (ROI) to correct for
residual decoupled prostate motion. Finally, a B-spline deformable registration with
a grid size of 5 × 5 × 5 was used to fine-tune the alignment and to adjust for tissue
deformation that occurred during the procedure. The whole registration process is
fully automatic after the initial user selection of the rectangular prism prostate ROI
on the fixed image by specifying the starting position and size.
Due to the aforementioned large differences in our images, mutual information was
chosen to be the similarity metric. The mutual information metric has a distinctive
property that it does not require a known mapping function between the intensities of
different images, but only assume the existence of a probabilistic relationship. Mutual
information (I) is the amount of information that one can obtain about a random
variable (X) by observing another random variable (Y ). The random variables in this
case are the image intensities. Mutual information can be expressed by the following
equations:
I(X, Y ) = H(X) + H(Y ) − H(X, Y )
(4.1)
I(X, Y ) = H(X) − H(X|Y ) = H(Y ) − H(Y |X)
XX
p(x, y)
I(x, y) =
p(x, y) log
p(x)p(y)
y∈Y x∈X
(4.2)
(4.3)
where H(X) and H(Y ) are the marginal entropies of X and Y respectively, and
H(X, Y ) is their joint entropy. H(X|Y ) is the conditional entropy of X given Y , and
H(Y |X) is the conditional entropy of Y given X. The joint probability density of
4.3. METHODS
44
Figure 4.2: Workflow of the three-stage registration algorithm between the pre- and
post-needle insertion volumes using mutual information.
(X, Y ) is denoted as p(x, y), and the marginal probability densities are p(x) and p(y).
Furthermore, our implementation involved a variant of the gradient descent optimizer for versor rigid 3D transform, and a L-BFGS-B (Limited-memory BroydenFletcher-Goldfarb-Shannon with simple bounds) optimizer for the deformable component.
4.3.3
Registration Validation
In the clinical MR images, the exact correspondence of the prostate anatomy cannot
be identified easily. In addition, prostate movement can be decoupled from surrounding organs and bony structures. Therefore, typical validation methods such as using
landmarks to evaluate the registration accuracy are not applicable in our case. To
validate our registration algorithm, we first generated simulated image volumes by
applying known transformations (ground truth) to an existing image volume. The
ground truth values were chosen based on observations of prostate dislocation from
4.3. METHODS
45
the images and prior expert knowledge. It was found that a range of ± 20 mm translation and ± 10◦ rotation along each of the three axis was large enough to cover
majority of the prostate motion during biopsy. The target displacement differences
between the ground truths and recovered transformations generated by the algorithm
from registering simulated volumes with the original volume was calculated.
We then proceeded to validate the algorithm on actual clinical image pairs by
performing image overlays and evaluating the prostate contour alignment between
the resulting volumes with its corresponding fixed volume. The contours were also
confirmed with the radiologists involved in this study. This process was done in 3D
Slicer [67], a free open source software package for visualization and image analysis.
All of the images in the dataset were verified for its registration accuracy based on the
image overlay method. If the maximum contour misalignment (measured using the
RAS physical space feature in 3D Slicer) after the proposed three-stage registration
is over 2 mm, then manual registrations were performed. We set the 2 mm standard
for registration accuracy because the MR image spacing of the slices is 3 mm, and
therefore it can ensure the correct alignment of the image slices within the volumes.
4.3.4
Biopsy Accuracy Analysis
To evaluate quantitatively and analyze the biopsy accuracy, we defined and studied
the following three terms (Figure 4.3):
ˆ Target displacement: the distance between planned (pre-needle insertion) and
actual (post-needle insertion) biopsy target. The actual target location was
obtained by applying the transformation from the registration algorithm to the
planned target (therefore it is dependent on the registration accuracy). To
4.3. METHODS
46
determine whether this dislocation is the same as the needle insertion direction,
the displacement was decomposed into two components: one parallel and one
orthogonal to the needle vector (Figure 4.4). A Wilcoxon signed rank test
(non-parametric test for paired samples) was conducted to see whether target
movement in the needle direction was significantly higher than the orthogonal
direction.
ˆ Needle placement error: the distance from the planned biopsy target to the
biopsy needle trajectory line. This distance indicates how much the robot had
missed the intended target, assuming no prostate motion during the biopsy
procedure. The needle trajectory line was obtained using two needle tip coordinates from the post-insertion volume. Commonly used titanium needles
are not directly visible in MRI, but they generate an artifact in the immediate neighborhood of the needle. Therefore, the true needle position may differ
from the artifact position. However, in this particular case, the needle artifact
errors are significantly smaller than the errors due to patient motion and tissue
deformation [77].
ˆ Biopsy error: the distance from the actual biopsy target to the needle trajectory
line. This is the most relevant metric for assessing biopsy accuracy, since the
length of the tissue core excised by the needle is about 20 mm long; hence
target movement orthogonal to the needle trajectory is of our main concern. To
further study the orthogonal component of the displacement, it was separated
into RAS coordinates and principle component analysis (PCA) was performed
on the data.
4.3. METHODS
47
Figure 4.3: Illustration of the prostate dislocation during needle insertion and the
parameters used in biopsy accuracy analysis.
Figure 4.4: The decomposition of target displacement into parallel and orthogonal
component.
4.4. RESULTS
4.4
48
Results
4.4.1
Registration Accuracy
The patient data selection for this study simply requires available planning and needle confirmation image volumes along with the corresponding planned biopsy target
coordinates. A total of 90 image sets from 24 patients were studied.
The accuracy of the registration procedure was studied in order to provide a bound
on biopsy accuracy evaluation. Images from 5 randomly selected patients were each
transformed by a different ground truth. The target displacement differences between
all of the ground truth and the recovered transformations from the algorithm were less
than 1.0 mm, which demonstrates the correctness of the method to recover prostate
dislocation using simulated images. The actual clinical image pairs are quite different
from the simulated images since they were acquired with different imaging sequences.
Therefore, further testings with simulated images are not necessary.
The registration results from all 90 clinical image pairs were validated using the
previously discussed image overlay approach. Majority of the images registered well
(accurate to within 2 mm based on contour evaluations), and only a small subset
(22%) were over the 2 mm error. The inaccuracy from the automatic registration
was mainly due to poor image quality, such as the existence of large glares around
the endorectal coil that the bias correction was not able to recover. After manual
adjustments for these images, all registrations were accurate to within 2 mm. In this
study, all of the contour alignments were verified by the same person. However, it is
important to note that there are some intra-observer variability in prostate contouring
[84]. Figure 4.5 shows an example of the prostate contour overlay before and after the
automatic registration. The signed rank test has shown that the results from rigid
4.4. RESULTS
49
Figure 4.5: Prostate contour overlays before (left) and after (right) the automatic
registration.
and deformable registrations were significantly different (p = 4.6 × 10−4 ). However,
rigid registrations recovered the majority (88%) of the target displacement.
As part of our validation process, we also manually registered the rectum and
pubic bone from some of the images separately to verify if patient and robotic device motion were different from that of the prostate. We chose to estimate patient
motion by measuring the displacement of the pubic bone, and robotic motion by the
displacement of the rectum, since it contains the endorectal imaging probe of the
biopsy device. We found that the prostate motion was different from its surrounding structures, and it was more similar to the bone motion than the rectum motion.
This confirmed our assumption that prostate motion does differ from its surrounding
structures even though it can be affected by them, and this should be taken into
consideration when performing registration.
4.4. RESULTS
4.4.2
50
Biopsy Accuracy
The mean, range, and standard deviation for target displacement, needle placement
error, and biopsy error are summarized in Table 4.1. The histograms of these measurements for all 90 biopsies are shown in Figure 4.6 a-c. Furthermore, target displacements in RAS coordinates are plotted in Figure 4.6 d-f. Lilliefors normality tests
(for unknown distribution mean and variance of small samples) were conducted and it
was found that none of the target displacements, needle placement errors, and biopsy
errors are normally distributed.
The parallel (mean 3.1 mm) and orthogonal (mean 3.6 mm) component of the
displacement to the needle trajectory was computed and found to be not statistically
different (p = 0.3) from one another, based on a signed rank test. For the parallel
component, only 32% of the targets moved toward the needle insertion direction
(mean 2.8 mm), and the rest 68% went in the opposite direction (mean 3.3 mm).
Since the biopsy tissue core is about 20 mm in length, it was still able to excise the
tissue that had displaced in the direction that is parallel to the needle. A PCA was
performed on the orthogonal component in RAS coordinates. The resulting first two
principal components ([1, 0.1, -0.1] and [-0.2, 0.9, -0.4]) accounted for 96% of the data
variance.
To study the effect of patient movement on biopsy accuracy, 22 biopsies that contained lateral patient motion (determined by visual inspection of the displacement
of rectum and pubic bone in 3D Slicer) greater than 5 mm were grouped separately.
Motions larger than 5 mm were suspected to be caused by involuntary patient movements such as pelvis movement or reflexive muscle clenching due to discomfort, which
is in nature different from prostate dislocation and deformation caused by the needle.
4.4. RESULTS
51
Figure 4.6: Left: Histograms of target displacement (top), needle placement errors
(middle), and biopsy errors (bottom) of the 90 biopsies. Right: Transverse
(top), sagittal (middle), and coronal (bottom) view of the prostate target
displacement. ’*’ and ’.’ represent biopsies on the left and right prostate
lobes, respectively.
4.5. DISCUSSION
52
Table 4.1: Data statistics for accessing biopsy accuracy
Mean
Range
Target
Displacement
(mm)
5.2 4.4* 7.8**
0.9- 0.91.618 14.8* 18**
3.5
3*
3.8**
Needle
Placement
Error (mm)
2.5
0.1-10.7
Biopsy
Error (mm)
4.3
0.2-12
3.8*
0.211.6*
2.8*
5.9**
1.712**
2.8**
Standard
1.6
2.9
Deviation
* Biopsies without the patient movement >5 mm group. ** Biopsies only for
patient movement >5 mm group.
The results for these 22 biopsies alone and the rest of the 68 biopsies are listed in
Table 4.1. The large patient motion caused a 2.1 mm increase in the mean biopsy
error. Biopsies performed at the left and right side of the prostate were also analyzed
separately. Results show that 43% of right biopsies had a prostate displacement towards right, and 61% of left biopsies had displacement toward left. This is because
the clinicians may have introduced a systematic bias by always using the beveled tip
needles in the same orientation.
4.5
Discussion
The results from our three-stage registration algorithm allowed for quantitative evaluation of the targeting accuracy for the APT-MRI system as well as prostate motion
analysis during biopsy. A clinically significant tumor has a minimum volume of 0.5
cm3 [69], which corresponds to a sphere with a radius of approximately 5 mm. Therefore, the maximum error should be less than 5 mm to not miss the targeted region.
The targeting accuracy of the APT-MRI system is considered to be acceptable, since
the mean of its needle placement error is 2.5 mm (Table 4.1). This implies that the
4.5. DISCUSSION
53
robotic device was accurate enough to place the needle at the intended biopsy target
assuming no prostate movement during the procedure. However, the prostate did
dislocate and deform upon needle insertion. Based on the 90 biopsy cases used in
this study, the mean prostate displacement was over 5 mm. This resulted in a mean
biopsy error of 4.3 mm. Furthermore, 28% of the biopsies have an error greater than
5 mm, and this error is higher for cases with large patient motion (Table 4.1). To
monitor for gross and sudden changes to the prostate location due to patient motion,
real-time tracking by plane-to-volume registrations can be used [90].
The biopsy needle was inserted approximately in the superior-anterior direction
towards the prostate. However, statistical test indicated that only half of the prostate
dislocation was in this direction. The majority of the variance from the other half
can be captured by the first two eigenvectors from PCA of the orthogonal component
of the displacement. This implies that the component of prostate dislocations that
caused the majority of the biopsy errors can be defined by a plane with vectors [1,
0.1, -0.1] and [-0.2, 0.9, -0.4]. It is important to note that the biopsy needles have a
beveled tip, which can also have an effect on the biopsy target dislocation.
The separate registrations of prostate, rectum, and pubic bone indicated that the
prostate was capable of moving independently of its surrounding structures. This confirms with our assumption, and further validates our muti-step registration approach
that uses both the entire image and only the prostate as ROI. The reason why its
movement was more similar to the pubic bone may be due to the transrectal robotic
device limiting rectum movement when the patient moves. The registration results
also have shown that even though the majority of prostate motion during biopsy is
4.5. DISCUSSION
54
rigid, there was also a significant amount of deformation caused by the needle insertion process. Therefore, to track the precise target location during biopsy, deformable
registration is recommended in addition to rigid registration.
In conclusion, a retrospective accuracy analysis of an MRI-guided robotic prostate
biopsy system [51] [50] was performed on 90 image sets by using a three-stage registration procedure to capture prostate motion during biopsy. The volumetric and soft
tissue imaging capabilities of MRI enabled us to identify the needle location in relation
to the prostate anatomy. In addition, the registration results allowed for quantitative
characterization of prostate dislocation and deformation during transrectal biopsy.
It was found that majority of the prostate motion during the procedure were rigid,
but there were also a significant amount of deformation during the process. Furthermore, the prostate moved differently from its surrounding structures. The exact
amount of this motion and deformation cannot be determined without fiducials or
finer volume images. However, even taking into account of the imperfections of our
validation framework, the results demonstrate that there is a substantial amount of
biopsy errors that should not be ignored. Further research on prostate motion and deformation upon needle insertion should be conducted to facilitate the development of
motion compensation techniques, which can be incorporated into the clinical protocol
to increase biopsy accuracy.
55
Chapter 5
Target Localization in MRI-guided Transperineal
Prostate Biopsy using Multi-Slice-to-Volume
Registration
At the time of writing this thesis, this chapter was prepared for journal submission.
5.1
Overview
Prostate needle biopsy is a commonly performed procedure since it is the most definitive form of cancer diagnosis. Magnetic resonance imaging (MRI) allows targetspecific biopsies to be performed. However, needle placements are often inaccurate
due to intra-operative prostate motion and the lack of motion compensation techniques.
Purpose: To detect and determine the extent of tissue dislocation during an MRIguided biopsy so that the needle insertion plan can be adjusted accordingly.
Method: A multi-slice-to-volume registration algorithm was developed to align the
pre-operative planning image volume with three intra-operative orthogonal image
slices of the prostate acquired immediately before needle insertion. The algorithm
5.2. INTRODUCTION
56
consists of an initial rigid transformation followed by a deformable step.
Results: A total of 14 image sets from 10 patients were studied. Based on prostate
contour alignment, the registrations were accurate to within 2 mm.
Conclusion: This algorithm can be used to increase the needle targeting accuracy
by alerting the clinician if the biopsy target has moved significantly prior to needle
insertion. The proposed method demonstrated feasibility of intra-operative target
localization and motion compensation for MRI-guided prostate biopsy.
5.2
Introduction
Current clinical biopsy protocols do not take into consideration of possible intraoperative prostate motion and deformation during the procedure due to both patient
movement and mechanical forces exerted by the biopsy system. The prostate is a soft
tissue organ attached only by connective tissues. Therefore, it can shift, rotate, and
deform differently from the surrounding structures. Biomechanical modeling of the
organ behaviour during biopsy is extremely complex since its parameters and material
properties can vary between patients. Our previous study [22] found a mean prostate
centroid motion of 8.7 mm (range: 0.2-34.7 mm) during MRI-guided transperineal
biopsies based on 538 images. Xu et al. [109] reported a mean prostate displacement
of 5.2 mm (range: 0.9-18 mm) for 90 needle insertions in MRI-guided transrectal
biopsy, and 28% of the biopsy errors exceeded 5 mm, which corresponds to the radius
of a clinically significant tumor (0.5 cc) [69].
Intra-operative prostate motion and deformation can cause inaccurate needle
placement during biopsy [109]. As a result, malignant tumors can be missed, which in
turn will lead to an increased number of repeated biopsies and delaying of treatment.
5.2. INTRODUCTION
57
Therefore, it is important to incorporate motion tracking techniques into the clinical
procedure in order to improve the overall needle targeting accuracy.
The ultimate goal would be to track both the intended biopsy target and the
needle in real-time. However, as the first step, we focus on estimating the target
position and developing a motion detection method to warn the clinician if there is
large dislocation of the intended biopsy target prior to needle insertion. If possible,
the clinician can then adjust the needle insertion plan to compensate for this motion.
Otherwise, a new planning scan can be acquired, and the insertion plan can be made
based on this new image. Such methods need to be integrated into imaging protocols
available on any regular MRI scanner. Since volumetric MRI (multi-slice 2D imaging
series) acquisition typically takes around a minute, methods that require multiple
of these image series to locate the biopsy targets are not ideal. As an alternative,
intermittent acquisition of only a few image slices can be used to obtain the required
information of a full image volume. This process takes considerably less time, since
a single image slice can usually be obtained in only a few seconds.
5.2.1
Related Works
There are numerous methods developed for tracking the prostate during biopsy. However, most of these methods use ultrasound images or focus on ultrasound and MRI
fusion. In the context of MR image-based prostate registration using multiple image
slices, there are three closely relevant papers by Fei et al. [23], Gill et al. [34], and
Tadayyon et al. [90].
Fei et al. developed a single-slice-to-volume rigid registration algorithm for radiofrequency thermal ablation of prostate cancer. The pre-operative image volumes and
5.2. INTRODUCTION
58
intra-operative image slices in transverse, sagittal, and coronal planes were acquired
from 3 volunteers using a conventional 1.5T scanner and a clinical 0.2T C-arm open
MRI scanner, respectively. The image slices from each volunteer were individually
registered to the corresponding MRI volumes. A total of 450 registrations were performed and it was found that images in the transverse orientation produced the best
results based on comparison with volume-to-volume registration. The algorithm used
two similarity metrics and featured a multi-resolution approach with an automatic
restart. The restart applied a random perturbation to the last transformation parameters found by the registration in order to to escape the potential local optima of
the cost function.
Gill et al. addressed the problem of local extremes and the inefficiency in Fei’s
optimization. The need for restarting the registration was eliminated by using a multiresolution registration based on a region of interest. Due to insufficient information
from a single transverse image slice, Gill’s implementation incorporated a simulated
sagittal slice centered at the prostate to improve the registration result.
Both studies mentioned above do not take into consideration of prostate deformation, which is a known issue during biopsy. Tadayyon et al. proposed a non-rigid
method to account for prostate deformation in addition to rigid motion. The algorithm can handle multiple image slices with different orientations all at once without
using the multi-resolution scheme. Simulated intra-operative image slices were pasted
into an empty volume (sparse volume construction) and 3D-to-3D image-based registration were performed. The study showed that three image slices in the transverse,
sagittal, and coronal plane were sufficient enough to be used for image registration
purposes. It is important to note that this exploratory study only used simulated
5.2. INTRODUCTION
59
image slices based on the high resolution image volume. Real clinical intra-operative
image slices can be quite different to these simulated images since a different imaging
protocol is often used. Furthermore, the mean execution time for the registrations
was over 16 minutes, which is not clinically practical.
This chapter reports a multi-slice-to-volume deformable registration method that
aligns the pre-operative planning image volume with three intra-operative orthogonal
image slices acquired immediately before needle insertion. The method was tested on
clinical images provided by Brigham and Women’s Hospital (BWH) in Boston, MA,
United States. We present several major improvements to Tadayyon’s work. These
include:
ˆ Method validation on actual clinical data instead of just simulated images
ˆ Ability to handle image slices in any orientation rather than just transverse,
sagittal, and coronal
ˆ Elimination of the sparse volume construction step, and register the image slices
directly to the planning volume without any resampling
ˆ Improved method validation techniques
ˆ Region of interest is no longer needed in the rigid step
ˆ Incorporation of image pre-processing for MRI bias field correction
The remainder of this paper presents the detailed methodology and validation of
our multi-slice-volume registration using clinical images from MRI-guided transperineal prostate biopsy.
5.3. METHODS
5.3
60
Methods
The objective is to detect cases where large intra-operative motion is present prior
to needle insertion so that either modifications of the original biopsy plan can be
made (compensating for the motion) or re-acquisition of the planning volume can
be performed to compensate for the target dislocation. By registering the intraoperative orthogonal image slices with the pre-operative planning volume, the existence of prostate motion and deformation can be determined.
5.3.1
Image Acquisition
A custom setup and software was developed at BWH to perform MRI-guided transperineal biopsy without moving the patient out of the scanner [98]. The setup consists
of a specially designed tabletop and a needle guiding template that gives clinicians
access to the perineum of the patient at the imaging position.
The remainder of this section describes the portion of the clinical protocol that
is relevant to the acquisition of the images used in this study. First, the patients
were sedated and immobilized on the table top with velcro wraps. Then they were
placed into a wide bore Siemens Magnetom Verio 3T scanner in supine position,
and a 4-minute T2-weighted multi-slice 2D transverse imaging series covering the
whole prostate gland were taken with the turbo spin-echo (TSE) sequence. This
was used as the pre-operative planning volume where potential biopsy targets in
scanner coordinate system were selected by the clinicians. Immediately before the
needle insertion, a quick 18-second scan of 3 orthogonal image slices in the transverse,
sagittal, and coronal plane were collected at approximate center of the prostate with
half-Fourier acquisition single-shot turbo spin-echo (HASTE) localizer sequence. The
5.3. METHODS
61
Figure 5.1: Clinical workflow of the image acquisitions during MRI-guided prostate
biopsy. The solid lines indicate the standard biopsy procedure, and the
dashed lines represent the additional orthogonal image slice acquisition
specific to our study. The grey box is the motion compensation registration method we propose to incorporate.
clinicians then proceeded with the needle placement. Finally, another multi-slice
transverse T2-weighted sequence (1-minute scan) were acquired with the needles in
place to confirm its placement. Figure 5.1 illustrates the overall clinical workflow.
The process starting from the initial acquisition of the pre-operative planning
volume to the final needle confirmation image takes an average of 90 minutes. During
this time, intra-operative motion of the prostate may occur. If this motion is large,
then the biopsy target locations that were chosen based on the pre-operative planning
volume would no longer be valid.
5.3. METHODS
5.3.2
62
Image Registration
The images used in the multi-slice-to-volume registration were the pre-operative
planning T2-weighted volume (multi-slice 2D transverse image series) and the intraoperative orthogonal image slices. All images were first pre-processed to correct for
nonuniform intensity caused by field inhomogeneities. Since the biopsy procedure at
BWH did not include the use of an endorectal coil, the pre-processing step does not
have a large effect on their images. However, this step is included in the algorithm so
that it can be applied to a broader range of images which are taken with the endorectal
probe. N4ITK (Nicks N3 Insight Toolkit) implementation for MRI bias field correction [99] was used because it does not require expert supervision, user interaction,
or training, and only has a few user-defined parameters. The two most important
parameters are bias full width at half maximum (BWHM) and noise. BWHM defines
the Gaussian that estimates the bias field, and noise specifies the Wiener filter used
for field estimation. Values of 0.5 for BWHM and 0.01 for noise were found to work
the best with our clinical images.
After bias field correction, a multi-slice-to-volume registration algorithm (Figure
5.2) was developed using ITK [115] to determine the transformation between the
pre-operative planning volume (V ) and intra-operative orthogonal image slices (S )
acquired just prior to needle insertion. The fixed and moving images of the registration were S and V , respectively. Due to imaging protocol differences, the same
tissue structures have different intensities on the fixed and moving images. Therefore,
mutual information was chosen as the metric for evaluating image alignment instead
of metrics such as cross correlation that uses direct comparison of pixel intensities.
Since each of the three orthogonal image slices needs to be registered to the planning
5.3. METHODS
63
volume, the mutual information metric was modified so that it calculates the sum of
these three metrics.
The algorithm consists of an initial rigid registration using the entire image to
correct for gross prostate motion in coherence with the device and patient. To recover
tissue deformation, a B-spline deformable registration with grid size of 5 × 5 × 5 was
performed using only the prostate as the region of interest. The B-spline transform
is able to represent a typical prostate deformation. In addition, it is fast to compute,
and the number of grid points and its maximum dislocation can be used to keep the
deformation field under control during registration. A gradient descent optimizer was
implemented for the versor rigid 3D transform, and an L-BFGS-B (Limited-memory
Broyden-Fletcher-Goldfarb-Shannon with simple bounds) optimizer was used for the
deformable component. The detailed workflow is shown in Figure 5.2. The entire
process is fully automatic after the initial user selection of the rectangular prism
prostate ROI on the fixed image by specifying the starting position and size. One of
the main implementation challenges was to tune a set of registration parameters that
would work the best for our imaging sequence. These parameters were chosen based
on both speed and accuracy of the registration result.
5.3.3
Registration Validation
Experiment 1 - Simulated Images
To validate the accuracy of rigid registration, we first simulated intra-operative orthogonal image slices from the pre-operative planning volume. Three empty image
slices in transverse, sagittal, and coronal plane were created, and the pixel information from the planning volume were resampled into the slices. These simulated images
5.3. METHODS
64
Figure 5.2: Workflow of the multi-slice-to-volume registration between the preoperative planning volume and intra-operative image slices.
were then registered back to the planning volume with a random initial transform that
was set to a range of ± 20 mm translation and ± 10◦ rotation along each of the three
axes. If the rigid registration works properly, it should correct for this initial misalignment and produce a transformation matrix that is close to the identity. Five points
in the center, left, right, top, and bottom of the prostate (Figure 5.3) were chosen on
the fixed images, and its target registration error (TRE) were calculated. The TRE
was defined as the distance between one of the chosen points and its reconstructed
point. The process was repeated 5 times for each of the 11 planning images we have,
and all of the TREs with its corresponding initial misalignment were recorded.
5.3. METHODS
65
Figure 5.3: The arrangement of 5 ground truth points chosen on the prostate.
The accuracy of deformable registration was also tested using simulated intraoperative orthogonal image slices. We selected two corresponding sets of five points
on the planning volume. The second set of points were randomly displaced ± 5
mm in plane from the first set. A Landwarp landmark deformable registration by
Plastimatch was performed in 3D Slicer [67], a free open source software package for
visualization and image analysis. The planning volume was warped based on the
displacements of the two point sets using thin-plate spline transform and radial basis
functions [79]. The reason we chose to warp the planning volume instead of the image
slices is simply because the Landwarp landmark deformable registration works better
with volumetric images. The three orthogonal image slices were then generated using
the warped volume, and were registered to back to the original planning volume using
the B-spline deformable algorithm that we developed. The displacement vectors of
the five points were computed and compared with the ground truth. The process was
repeated once for each of the 11 planning images, and all TREs were calculated and
recorded as an accuracy assessment.
The numerical ranges for the ground truth initial rigid misalignment and warping
were chosen based on our previous study of prostate motion during MRI-guided biopsy
[109]. It was found that the majority (88%) of the intra-operative motion were rigid
and the amount of tissue deformation was small (mean: 0.7 mm, max: 4.9 mm).
5.3. METHODS
66
Therefore, a range of ± 20 mm translation and ± 10◦ rotation along each of the three
axes, and ± 5 mm control points displacement for image warping are large enough to
cover most of the intra-operative prostate behaviour.
Experiment 2 - Clinical Images
For the clinical images, a different validation method was applied since the exact correspondence of the prostate anatomy (calcifications or tissue morphology) between
different images cannot be easily identified. The fixed and moving images were overlaid in 3D Slicer before and after registration, and the prostate contour alignment were
examined. After the contour alignment validation, the transformations produced by
the algorithm were used as the ground truth. The above validation procedures used
on simulated images were repeated on the actual clinical images to test the robustness
of the registration.
Experiment 3 - Multi-slice-to-volume vs. Volume-to-volume
To determine whether the multi-slice-to-volume registration is sufficient in capturing
prostate motion intra-operatively, we compared it with volume-to-volume registration.
Since volumetric images contain much more information than the three orthogonal
slices, its registration results can be used as a reference to evaluate the accuracy of the
multi-slice-to-volume registration. The needle confirmation image volume (Figure 5.1)
for each biopsy was registered with its corresponding planning volume using an initial
rigid alignment of the whole image, followed by a B-spline deformable transform with
only the prostate as region of interest. This is similar to what Xu et al. [109] proposed
for the registration of pre- and post-needle insertion volumetric MRIs for transrectal
5.3. METHODS
67
Table 5.1: Registration validation experiments summary
Exp.
Registration
1a
Rigid,
MSV
1b
Deformable,
MSV
2a
Fixed
Image(s)
3 orthogonal slices
resampled from
planning volume
3 orthogonal slices
resampled from warped
planning volume
3 orthogonal
slices
3 orthogonal
slices
Moving
Image
Planning
volume
Planning
volume
Rigid,
Planning
MSV
volume
2b
Rigid+
Warped
deformable,
planning
MSV
volume
3a
Rigid+
3 orthogonal slices
Planning
deformable,
resampled from needle
volume
MSV
confirmation volume
3b
Rigid+
needle
Planning
deformable,
confirmation
volume
VV
volume
Abbreviations: Exp. = Experiment; MSV = Multi-Slice-to-Volume; VV =
Volume-to-Volume.
prostate biopsy. The results were compared to the multi-slice-to-volume registration
between the planning volume and simulated intra-operative orthogonal images slices
generated from the needle confirmation volume. The 5 points on the prostate (Figure
5.3) were selected on the fixed (needle confirmation) images, and were transformed
by the results obtained from both registrations separately. The distances between
target reconstructions from volume-to-volume registration and multi-slice-to-volume
registration were also calculated.
Table 5.1 summarizes all of the images and methods used for these three registration validation experiments.
5.4. RESULTS
68
Figure 5.4: Examples of clinical image prostate contour overlay in the transverse
plane. Each of the three images are copies of the same fixed image overlaid with the contours from the moving image a) before registration, b)
after rigid registration, c) after deformable registration.
5.4
Results
A total of 14 planning volume and orthogonal image slices pairs from 10 patients were
obtained from the Brigham and Women’s Hospital and were used in our study.
All resulting images after the multi-slice-to-volume registration were overlaid with
their corresponding fixed images in 3D Slicer. Based on prostate contour evaluations,
the misalignments were all under 2 mm (measured using the RAS physical space
feature in 3D Slicer). An example of the image overlay before and after registration
of the clinical image pairs is shown in Figure 5.4. The contours are all verified
with radiologists involved in the project. Even taking into account of the intraobserver variability in the contouring [84], the registration is still considered to be
fairly accurate.
The mean execution time for rigid and deformable registrations were 3.3 and 31
seconds, respectively, on an Intel Core i7-2600K processor running at 3.40GHz. This
is summarized in Table 5.2. Furthermore, the N4ITK bias field correction only takes
5.4. RESULTS
69
Table 5.2: Summary of execution time statistics for multi-slice-to-volume registration
with clinical images
Mean
Range
Standard deviation
Rigid
3.3 sec
2.5-4.2 sec
0.6 sec
Deformable
31 sec
13.7-68.1 sec
14.8 sec
Table 5.3: Gradient descent optimizer parameters for the rigid registration step.
Parameters
Max step length
Max step length
Relaxation factor
Max number of iterations
Value
2
0.01
0.9
250
less than 2 seconds to complete, and therefore does not have a major impact on the
overall computation time.
The difference between rigid and deformable registrations was studied. The results
were significantly different (p = 7.0 × 10−10 ), but rigid registration was able to recover
95% of the total prostate displacement on average.
Values of the optimizer parameters for both rigid and deformable steps are listed
in Table 5.3 and Table 5.4, respectively. For rigid registration, a variation of the
gradient descent optimizer was used. The maximum step length and relaxation factor
define the convergence rate and robustness of the algorithm. The minimum step length
and maximum number of iterations specify the stopping conditions. In the deformable
step, L-BFGS-B was used as the optimizer. The bounds parameter sets the maximum
amount of allowed deformation, while maximum number of corrections is used for
updating the limited memory Hessian matrix. Convergence factor, projected gradient
tolerance, and maximum number of iterations are the terminating conditions. Their
values were chosen based on both speed and accuracy of the results.
5.4. RESULTS
70
Table 5.4: L-BFGS-B optimizer parameters for the deformable registration step.
Parameters
Convergence factor
Projected gradient tolerance
Number of corrections
Max number of iterations
Upper and lower bound
Value
1 × 1012
1 × 10−7
10
250
±5
Table 5.5: TRE statistics for Experiment 1a: multi-slice-to-volume rigid registration
between the planning volume and simulated intra-operative orthogonal
image slices. The simulated images were undistorted (no deformation was
applied)
Mean
Range
Standard deviation
5.4.1
Rigid Registration TRE
0.1 mm
0-0.9 mm
0.1 mm
Registration Validation
Experiment 1 - Simulated Images
The accuracy of both rigid and deformable registrations were studied separatively
using simulated intra-operative orthogonal image slices. For rigid registration, a mean
TRE of 0.1 mm was found. All TREs from 55 simulations are shown in Figure 5.5,
and Table 5.5 provides a summary of the TRE statistics for the rigid step. For the
deformable part of the algorithm, the mean TRE was 0.5 mm. This is summarized in
Table 5.6. Figure 5.6 shows all TREs from the 11 experiments and their corresponding
initial misalignments. The prostate contour alignment before and after deformable
registration using the simulated images are shown in Figure 5.7.
5.4. RESULTS
71
Figure 5.5: TREs for Experiment 1a: the 55 multi-slice-to-volume rigid registrations
using simulated images. The simulated images were undistorted (no deformation was applied). A random initial transform of ± 20 mm translation
and ± 10◦ rotation along each axis were set to the registrations. The distance between the initial target and the ground truth target was defined
as the initial misalignment.
Table 5.6: TRE statistics for Experiment 1b: multi-slice-to-volume deformable registration between the planning volume and simulated intra-operative orthogonal image slices
Mean
Range
Standard deviation
Deformable Registration TRE
0.5 mm
0-1.6 mm
0.4 mm
5.4. RESULTS
72
Figure 5.6: TREs for Experiment 1b: the 11 multi-slice-to-volume deformable registrations using simulated images. The simulated images were warped
± 5 mm based on two sets of manually defined points using Landwrap
landmark deformable registration in 3D Slicer.
Figure 5.7: Examples of prostate contour overlay on the simulated fixed transverse
image slice a) before registration, b) after deformable registration.
5.4. RESULTS
73
Table 5.7: TRE statistics for Experiment 2a: multi-slice-to-volume rigid registration
between the planning volume and clinical intra-operative orthogonal image
slices
Mean
Range
Standard deviation
Rigid Registration TRE
0.04 mm
0-0.2 mm
0.1 mm
Table 5.8: TRE statistics for Experiment 2b: multi-slice-to-volume deformable registration between the planning volume and clinical intra-operative orthogonal image slices
Mean
Range
Standard deviation
Rigid Registration
Step TRE
3.9 mm
1-6.9 mm
1.7 mm
Deformable Registration
Step TRE
1.1 mm
0.3-2.6 mm
0.6 mm
Experiment 2 - Clinical Images
Using the original registration results from the actual clinical image pairs as the
ground truth, we further tested the accuracy and robustness of the algorithm with
only clinical images. In the rigid step, a random initial transform of ± 20 mm
translation and ± 10◦ rotation along each of the three axes was set. A total of 70
registrations were performed on 14 image pairs, and the TREs were all below 0.2 mm.
Table 5.7 and Figure 5.8 contains more detailed information regarding TRE statistics
for the rigid step. In the deformable step, the planning volumes were warped ± 5
mm and were then registered to the three orthogonal clinical image slices using the
two-step algorithm. The final mean TRE was found to be 1.1 mm for 14 registrations.
This is summarized in Table 5.8 and Figure 5.9. The rigid step recovered 93% of the
ground truth value, while the deformable alignment recovered 98%.
5.5. DISCUSSION
74
Figure 5.8: TREs for Experiment 2a: the 70 multi-slice-to-volume rigid registrations
using clinical images. A random initial transform of ± 20 mm translation
and ± 10◦ rotation along each axis were set to the registrations. The
distance between the initial target and the ground truth target was defined
as the initial misalignment.
Experiment 3 - Multi-slice-to-volume vs. Volume-to-volume
The results of multi-slice-to-volume registration between the needle confirmation images and the planning volumes were compared with volume-to-volume registration.
The target reconstruction distances of the two algorithms are plotted in Figure 5.10
(mean: 1 mm, range: 0.1-2.7 mm, standard deviation: 0.7 mm).
5.5
Discussion
The radius of a clinically significant tumor is 5 mm [69], hence a registration error less
than 5 mm is considered as sufficiently accurate for needle placement purposes. In
addition, it is also less than the slice spacing (3.6 mm) of our data set from Brigham
5.5. DISCUSSION
75
Figure 5.9: TREs for Experiment 2b: the 14 multi-slice-to-volume deformable registrations using clinical images. The image volumes were warped ± 5 mm
based on two sets of manually defined points using Landwrap landmark
deformable registration in 3D Slicer.
and Women’s Hospital.
It is important for the algorithm to be able to handle a large variety of MR images
with different intensities and initial misalignments. We performed several validation
methods to ensure the accuracy and robustness of our algorithm. Since the exact
correspondence of the prostate anatomy cannot be easily identified in clinical MR
images, typical validation methods such as using landmarks to evaluate the accuracy
of our registration are not applicable in this case. Therefore, we generated simulated
intra-operative orthogonal image slices from the pre-operative planning volume in
order to obtain ground truth.
5.5. DISCUSSION
76
Figure 5.10: Experiment 3: Histogram of the target reconstruction distances between
multi-slice-to-volume and volume-to-volume registration.
With regards to the actual clinical images, the image overlay after registration
showed obvious improvements to the initial contour alignment before registration
(Figure 5.4). The registration errors based on prostate contours were also clinically
acceptable (less than 5 mm). Since the patients were sedated, intra-operative patient
motion was limited. Therefore, the initial prostate misalignments before registration
were small for the Brigham and Women’s Hospital data. However, this is not the case
for transrectal biopsies. The registration results from both real and simulated images
demonstrated that our algorithm was able to recover initial rigid misalignments (±
20 mm translation and ± 10◦ rotation) and correct for prostate deformation (± 5
mm warping of the images) with less than 1.1 mm accuracy on average, which is well
under the 5 mm upper limit mentioned earlier. In an actual clinical setting, target
displacement that is parallel to the needle is less of a concern since the biopsy needle
core is about 20 mm long. Therefore, it is the orthogonal component of the target
5.5. DISCUSSION
77
displacement that contributes the most to the biopsy error.
The fact that our multi-slice-to-volume registration was able to produce results
that were less than 3 mm different from that of the volume-to-volume registration
further indicates the reliability of our algorithm.
Clinically, a processing time of approximately 1 minute to compute the current
biopsy target position prior to needle insertion is reasonable. The mean execution
time of 34.3 seconds fulfills this time requirement. The deformable step was much
more computationally expensive than the rigid step. Since rigid registration can capture 95% of the target displacement, it may also be tolerable to omit the deformable
step.
To summarize, we developed an image-based multi-slice-to-volume registration
algorithm for MRI-guided prostate biopsy to detect cases with large intra-operative
prostate motion prior to needle insertion. Contour alignments were used to validate
the registration results of clinical images, and both simulated and clinical images were
used for quantitative evaluation of the algorithm. All registration errors were well
below the radius of a clinically significant tumour (5 mm), and can be considered as
clinically acceptable. It was found that the majority of the prostate motion (95%)
during biopsy was rigid. The overall execution time for biopsy target dislocation
computation (34.3 seconds) was also short enough for clinical practice.
In conclusion, the multi-slice-to-volume registration can be used to alert the clinician if the biopsy target has moved significantly after the planning volume acquisition
and before needle insertion. Decisions can be made to either compensate the insertion
plan by the results of the slice-to-volume registration or to re-acquire another set of
planning images. The latter method takes more time and we hope in most cases it
5.5. DISCUSSION
78
will not be necessary and the registration result can be used for motion compensation
instead. Furthermore, a short image slice acquisition and registration time is needed
when a quick feedback loop is essential, such as in real-time image guided needle
steering. Currently, this method adds an extra step to the clinical protocol in order
to increase the needle targeting accuracy. In the future, needle tracking can also be
included to increase the speed of the procedure by reducing the need of the volumetric
needle confirmation imaging step. We expect further testing with more image data
and possible future integration of this methodology into clinical practice so that we
can make a tangible difference in the lives of patients.
79
Chapter 6
Conclusion
This chapter summarizes the main contributions and findings of this thesis, and proposes directions for future research.
6.1
Summary
A survey of the recent literature on the state-of-art in clinical image-guided prostate
needle placement procedures was presented in Chapter 2, and the prostate behaviour
upon needle insertion was reviewed in Chapter 3. Undeniably, research findings have
confirmed that there are a substantial amount of prostate motion and deformation
intra-operatively. Chapter 4 showed the effect of this motion and deformation on the
biopsy accuracy of the APT-MRI system. Finally, a motion compensation method
was proposed and validated in Chapter 5 to improve the needle targeting accuracy
for biopsy under MRI guidance.
6.1. SUMMARY
6.1.1
80
Prostate Motion and Deformation Upon Needle Insertion
Chapter 2 and 3 review the current image-guided needle placement procedures for
prostate cancer diagnosis and treatment. The extent of intra-operative prostate motion and deformation, and motion reduction methods used in the operating room were
included. The different kinds of procedures were organized based on their imaging
modalities and needle puncture path. The following is a list of the topics covered in
these chapters:
ˆ Types of imaging modalities for needle guidance
ˆ Needle puncture paths for prostate access
ˆ Patient positioning for different types of procedures
ˆ Patient-induced prostate motion
ˆ Prostate deformation due to imaging probes
ˆ Target dislocation during needle insertion and tissue acquisition
ˆ Motion reduction using special devices or equipment
ˆ Methods of intra-operative image registration
Overall, studies have shown that there is a substantial amount of prostate motion and deformation during image-guided needle placement procedures. Despite the
efforts to improve the basic TRUS-guided protocols, its image quality is a limiting
factor. MRI has been serving as an alternative to ultrasound for needle guidance.
However, due to its high cost and low availability, it is still at its early stages of
research.
6.1. SUMMARY
6.1.2
81
Accuracy Analysis in MRI-guided Robotic Prostate Biopsy
Knowing the possible intra-operative prostate motion and deformation during the
biopsy procedure, Chapter 4 aims to find out how much of the biopsy accuracy is
affected by this. The major contributions of this chapter include:
ˆ The development of a three-stage registration algorithm that aligns the pre-
and post-needle insertion image volumes to capture prostate motion and deformation during MRI-guided biopsy
ˆ Retrospective accuracy validation of the APT-MRI robotic biopsy system that
was used at the NCI for over 6 years
ˆ Quantitative analysis of the biopsy target dislocation during MRI-guided tran-
srectal biopsy under the APT-MRI system
The results have shown that even though the robotic needle placement is clinically
acceptable, there was still an average biopsy error of 4.3 mm due to intra-operative
prostate motion and deformation. It was also found that the majority of the prostate
motion during biopsy was rigid, and the prostate moved differently from its surrounding structures. Therefore, the incorporation of motion compensation techniques are
recommended in order to improve the overall biopsy needle targeting accuracy.
6.1.3
Target Localization in MRI-guided Transperineal Prostate Biopsy
using Multi-Slice-to-Volume Registration
Chapter 5 focused on developing a method to alert the clinicians if the biopsy target
has moved significantly prior to needle insertion. The insertion plan can then be
6.2. FUTURE WORK
82
adjusted accordingly in order to increase the needle targeting accuracy. The two
major contributions of this chapter include:
ˆ The development of a multi-slice-to-volume registration algorithm that aligns
the pre-operative image volume with 3 intra-operative image slices for biopsy
target localization
ˆ Accuracy validation of the algorithm on both clinical and simulated MR images
All quantitative and qualitative validation results of the registration algorithm
were clinically acceptable since the registration errors were well all below the radius
of a clinically significant tumour. In addition, the average execution time of the registration was also short enough to be clinically compatible. The multi-slice-to-volume
registration results were comparable with that of the volume-to-volume registration,
and therefore further confirms the accuracy of the algorithm. It was also found that
rigid registration can capture the majority of the target dislocation, which is consistent with previous findings.
6.2
Future Work
The actual clinical integration of the proposed multi-slice-to-volume registration for
motion compensation would require further testing with more clinical MR images
and the development of a user-friendly software interface. To significantly decrease
the procedure time, needle tracking can also be implemented so that the need of the
volumetric needle confirmation imaging step can be reduced. The ultimate goal is to
track both the intended biopsy target and the needle tip in real-time. The algorithm
can also be enhanced so that it can be applied to any needle placement procedures
6.2. FUTURE WORK
83
instead of just biopsy. In the long term, the software can even be further extended
beyond prostate cancer to other organ systems, especially for gynecology.
BIBLIOGRAPHY
84
Bibliography
[1] R. Alterovitz, K. Goldberg, J. Pouliot, I.J. Hsu, Y. Kim, S.M. Noworolski,
and J. Kurhanewicz. Registration of MR prostate images with biomechanical
modeling and nonlinear parameter estimation. Med. Phys., 33(2):446–454, 2006.
[2] D. Amsellem-Ouazana, P. Younes, S. Conquy, M. Peyromaure, T. Flam,
B. Debr, and M. Zerbib. Negative prostatic biopsies in patients with a high
risk of prostate cancer: is the combination of endorectal MRI and magnetic resonance spectroscopy imaging (MRSI) a useful tool? a preliminary study. Eur.
Urol., 47(5):582586, 2005.
[3] A.G. Anastasiadis, M.P. Lichy, U. Nagele, M.A. Kuczyk, A.S. Merseburger,
J. Hennenlotter, S. Corvin, K.D. Sievert, C.D. Claussen, A. Stenzl, and H.P.
Schlemmer. MRI-guided biopsy of the prostate increases diagnostic performance in men with elevated or increasing psa levels after previous negative trus
biopsies. Eur. Urol., 50(4):738–748, 2006.
[4] M. Baumann, P. Mozer, V. Daanen, and J J. Troccaz. Prostate biopsy tracking
with deformation estimation. Med. Image Anal., 16(3):562–576, 2012.
BIBLIOGRAPHY
85
[5] J. Bax, D. Cool, L. Gardi, K. Knight, D. Smith, J. Montreuil, S. Sherebrin,
C. Romagnoli, and A. Fenster. Mechanically assisted 3D ultrasound guided
prostate biopsy system. Med. Phys., 35(12):5397–5410, 2008.
[6] D. Beyersdorff, A. Winkel, B. Hamm, S. Lenk, S.A. Loening, and M. Taupuitz.
MR imagingguided prostate biopsy with a closed MR unit at 1.5 T: initial
results. Radiology, 234:576581, 2005.
[7] A. Bharatha, M. Hirose, N. Hata, S.K. Warfield, M. Ferrant, K.H. Zou,
E. Suarez-Santana, J. Ruiz-Alzola, A. D’Amico, R.A. Cormack, R. Kikinis,
F.A. Jolesz, and C.M. Tempany. Evaluation of three-dimensional finite elementbased deformable registration of pre- and intraoperative prostate imaging. Med
Phys., 28(12):2551–2560, 2001.
[8] P. Blumenfeld, N. Hata, S. DiMaio, K. Zou, S. Haker, G. Fichtinger, and
C. Tempany. Transperineal prostate biopsy under magnetic resonance image guidance: a needle placement accuracy study. J. Magn. Reson. Imaging,
26(3):688–694, 2007.
[9] K.K. Brock. Results of a multi-institution deformable registration accuracy
study (MIDRAS). Int. J. Radiation Oncology Biol. Phys., 76(2):583596, 2010.
[10] K.K. Brock, A.M. Nichol, C. Menard, J.L. Moseley, P.R. Warde, C.N. Catton, and D.A. Jaffray. Accuracy and sensitivity of finite element model-based
deformable registration of the prostate. Med. Phys., 35(9):4019–4025, 2008.
BIBLIOGRAPHY
86
[11] J.Y. Cho, S.H. Kim, and S.E. Lee. Peripheral hypoechoic lesions of the prostate:
evaluation with color and power Doppler ultrasound. Eur. Urol., 37(4):443–448,
2000.
[12] D. Citrin, H. Ning, P. Guion, G. Li, R.C. Susil, R.W. Miller, E. Lessard,
J. Pouliot, X. Huchen, J. Capala, C.N. Coleman, K. Camphausen, and
C. Menard.
Inverse treatment planning based on MRI for HDR prostate
brachytherapy. Int. J. Radiation Oncology Biol. Phys., 61(4):12671275, 2005.
[13] D.W. Cool, J. Bax, C. Romagnoli, A. Ward, L. Gardi, V. Karnik, J. Izawa,
J. Chin, and A. Fenster. Fusion of MRI to 3D TRUS for mechanically-assisted
target prostate biopsy: system design and initial clinical experience. In Prostate
Cancer Imaging, volume 6963, pages 121–133. MICCAI, LNCS, 2011.
[14] S.C. Cunico, A. Moroni, and G.M. Simeone. Imaging of Urogenital Diseases.
Springer-Verlag Italia, 2009.
[15] A. D’Amico, R. Cormack, S. Kumar, and C.M. Tempany. Real-time magnetic
resonance imaging-guided brachytherapy in the treatment of selected patients
with clinically localized prostate cancer. J. Endourol., 14(4):367–370, 2000.
[16] A.V. DAmico, C.M. Tempany, R. Cormack, N. Hata, M. Jinzaki, K K. Tuncali,
M. Weinstein, and J.P. Richie. Transperineal magnetic resonance image guided
prostate biopsy. J. Urol., 164:385387, 2000.
[17] M. Dattoli and K. Waller. A simple method to stabilize the prostate during
transperineal prostate brachytherapy. Int. J. Radiat. Oncol. Biol. Phys., 38:341–
342, 1997.
BIBLIOGRAPHY
87
[18] J. Dinkel, C. Thieke, C. Plathow, P. Zamecnik, H. Prm, P.E Huber, H.U. Kauczor, H.P. Schlemmer, and C.M. Zechmann. Respiratory-induced prostate motion: characterization and quantification in dynamic MRI. Strahlenther Onkol.,
187(7):426–432, 2011.
[19] H. Elhawary, Z.T. Tse, M. Rea, A. Zivanovic, B.L. Davies, C. Besant, N.M.
Desouza, D. McRobbie, I. Young, and M.U. Lamperth. Robotic system for
transperineal biopsy of the prostate: real-time guidance under MRI. IEEE
Eng. Med. Biol. Mag., 29(2):78–86, 2010.
[20] D.G. Engehausen, K. Engelhard, S.A. Schwab, M. Uder, S. Wach, B. Wullich,
and F.S. Krause. Magnetic resonance image-guided biopsies with a high detection rate of prostate cancer. Scientific World Journal, doi:10.1100:975971,
2012.
[21] K. Engelhard, H.P. Hollenbach, B. Kiefer, A. Winkel K. Goeb, and D. Engehausen. Prostate biopsy in the supine position in a standard 1.5-T scanner
under real time MR-imaging control using a MR-compatible endorectal biopsy
device. Eur. Radiol, 16:12371243, 2006.
[22] A. Fedorov, K. Tuncali, T. Penzkofer, J. Tokuda, S. Song, N. Hata, and C. Tempany. Quantification of intra-procedural gland motion during transperineal
MRI-guided prostate biopsy. In International Society for Magnetic Resonance
in Medicine (ISMRM) 21st Annual meeting, 2013.
[23] B. Fei, J.L. Duerk, D.T. Boll, J.S. Lewin, and D. L. Wilson. Slice-to-volume
registration and its potential application to interventional MRI-guided radiofrequency thermal ablation of prostate cancer. IEEE Trans. Med. Imaging,
BIBLIOGRAPHY
88
22(4):515–525, 2003.
[24] B. Fei, C. Kemper, and D.L. Wilson. A comparative study of warping and
rigid body registration for the prostate and pelvic MR volumes. Comput. Med.
Imaging Graph., 27:267–281, 2003.
[25] B. Fei, A. Wheaton, Z. Lee, J.L. Duerk, and D.L. Wilson. Automatic MR
volume registration and its evaluation for the pelvis and prostate. Phys. Med.
Biol., 47(5):823–838, 2002.
[26] V. Feygelman, J.L. Friedland, R.M. Sanders, B.K. Noriega, and J.M. Pow-Sang.
Improvement in dosimetry of ultrasound-guided prostate implants with the use
of multiple stabilization needles. Med. Dosim., 21(2):109112, 1996.
[27] G. Fichtinger, J. Fiene, C.W. Kennedy, G. Kronreif, I. Iordachita, D.Y. Song,
E.C. Burdette, and P. Kazanzides. Robotic assistance for ultrasound guided
prostate brachytherapy. In MICCAI, volume 4791, pages 119–127. SpringerVerlag Berlin Heidelberg, 2007.
[28] F. Frauscher, A. Klauser, H. Volgger, E.J. Halpern, L. Pallwein, H. Steiner,
A. Schuster, W. Horninger, H. Rogatsch, and G. Bartsch. Comparison of
contrast enhanced color Doppler targeted biopsy with conventional systematic
biopsy: impact on prostate cancer detection. J. Urol., 167(4):16481652, 2002.
[29] M. Fuchsjger, A. Shukla-Dave, O. Akin, J. Barentsz, and H. Hricak. Prostate
cancer imaging. Acta Radiol., 49(1):107–120, 2008.
BIBLIOGRAPHY
89
[30] D.B. Fuller, H. Jin, J.A. Koziol, and A.C. Feng. CTultrasound fusion prostate
brachytherapy: A dynamic dosimetry feedback and improvement method. a
report of 54 consecutive cases. Brachytherapy, 4:207–216, 2005.
[31] J.J. Futterer and J.O. Barentsz. MRI-guided and robotic-assisted prostate
biopsy. Curr. Opin. Urol., 22(4):316–319, 2012.
[32] M.C.G. Gbenou, A. Peltier, S.K. Addla, M. Lemort amd R. Bollens, D. Larsimont, T. Roumegure, C.C. Schulman, and R. van Velthoven. Localising prostate
cancer: comparison of endorectal magnetic resonance (MR) imaging and 3DMR spectroscopic imaging with transrectal ultrasound-guided biopsy. Urol.
Int., 88(1):12–17, 2012.
[33] M.P.R. Van Gellekom, M.A. Moerland, J.J. Battermann, and J.J.W. Lagendijk.
MRI-guided prostate brachytherapy with single needle methoda planning study.
Radiother. Oncol., 71(3):327332, 2004.
[34] S. Gill, P. Abolmaesumi, and S. Vikal. Intraoperative prostate tracking with
slice-to-volume registration in MR. In 20th Int. Conf. Soc. Med. Innov. Technol.,
pages 154–158, 2008.
[35] S.J. Haker, R.V. Mulkern, J.R. Roebuck, A.S. Barnes, S. DiMaio, N. Hata, and
C.M. Tempany. Magnetic resonance-guided prostate interventions. Top Magn.
Reson. Imaging, 16(5):35568, 2005.
[36] T. Hambrock, D.M. Somford, C. Hoeks, S.A.W. Bouwense, H. Huisman,
D. Yakar, I.M. Van Oort, J.A. Witjes, J.J. Ftterer, and J.O. Barentsz. Magnetic
BIBLIOGRAPHY
90
resonance imaging guided prostate biopsy in men with repeat negative biopsies
and increased prostate specific antigen. J. Urol., 183:520–528, 2010.
[37] M. Hirose, A. Bharatha, N. Hata, K.H. Zou, S.K. Warfield, R.A. Cormack,
A. DAmico, R. Kikinis, F.A. Jolesz, and C.M. Tempany. Quantitative MR imaging assessment of prostate gland deformation before and during MR imagingguided brachytherapy. Academic Radiology, 9(8):906–912, 2002.
[38] A.C. Hodge, A. Fenster, D.B. Downey, and H.M. Ladak. Prostate boundary
segmentation from ultrasound images using 2D active shape models: Optimisation and extension to 3D. Comput. Methods Programs Biomed., 84:99113,
2006.
[39] Y. Hu, T.J. Carter, H.U. Ahmed, M. Emberton, C. Allen, D.J. Hawkes, and
D.C. Barratt. Modelling prostate motion for data fusion during image-guided
interventions. IEEE Trans. Med. Imaging, 30(11):1887–1900, 2011.
[40] A.S. Huo, T. Hossack, J.L. Symons, R. Pebenito, W.J. Delprado, P. Brenner,
and P.D. Stricker. Accuracy of primary systematic template guided transperineal biopsy of the prostate for locating prostate cancer: a comparison with
radical prostatectomy specimens. J. Urol. 2012.
[41] E.I. Ibrahiem, T. Mohsen, A.M. Nabeeh, Y. Osman, I.A. Hekal, and M. Abou
El-Ghar. DWI-MRI: single, informative, and noninvasive technique for prostate
cancer diagnosis. Scientific World Journal, doi:10.1100:973450, 2012.
[42] S. Kadoury, P. Yan, S. Xu, N. Glossop, P. Choyke, B. Turkbey, P. Pinto,
B.J. Wood, and J. Kruecker. Realtime TRUS/MRI fusion targeted-biopsy for
BIBLIOGRAPHY
91
prostate cancer: a clinical demonstration of increased positive biopsy rates. In
A. Madabhushi et al., editor, Prostate Cancer Imaging, volume 6367, page 5262.
Springer-Verlag Berlin Heidelberg, 2010.
[43] I. Kaplan, N.E. Oldenburg, P. Meskell, M. Blake, P. Church, and E.J. Holupk.
Real time MRI-ultrasound image guided stereotactic prostate biopsy. Magn.
Reson. Imag., 20:295–299, 2002.
[44] V. Karnik, A. Fenster, J. Bax, D.W. Cool, L. Gardi, I. Gyacskov, C. Romagnoli,
and A.D. Ward. Assessment of image registration accuracy in three-dimensional
transrectal ultrasound guided prostate biopsy. Med. Phys., 37(2):802–813, 2010.
[45] Y. Kim, S. Noworolski, J. Pouliot, I. Hsu, and J. Kurhanewicz. Analysis of
prostate deformation due to different MRI/MRS endorectal coils for image fusion and brachytherapy treatment planning. Med. Phys., 31:1728, 2004.
[46] V.R. Kini, G.K. Edmundson, F.A Vicini, D.A Jaffray, G.Gustafson, and A.A
Martinez. Use of three-dimensional radiation therapy planning tools and intraoperative ultrasound to evaluate high dose rate prostate brachytherapy implants. Int. J. Radiation Oncology Biol. Phys., 43(3):571–578, 1999.
[47] A.P.S. Kirkham, M. Emberton, and C. Allen. How good is MRI at detecting
and characterising cancer within the prostate? Eur. Urol., 50:11631175, 2006.
[48] P. Koutrouvelis. Three-dimensional stereotactic posterior ischiorectal space
computerized tomography guided brachytherapy of prostate cancer: a preliminary report. J. Urol., 159(1):142–145, 1998.
BIBLIOGRAPHY
92
[49] P. Koutrouvelis, F. Hendricks, N. Lailas, G. Gil-Montero, J. Sehn, N. Khawand,
H. Bondy, and S. Katz. Salvage reimplantation in patient with local recurrent prostate carcinoma after brachytherapy with three dimensional computed
tomography-guided permanent pararectal implant. Technol. Cancer Res. Treat.,
2(4):339–344, 2003.
[50] A. Krieger, I. Iordachita, P. Guion, A.K. Singh, A. Kaushal, C. Menard, P.A.
Pinto, K. Camphausen, G. Fichtinger, and L.L. Whitcomb. An MRI-compatible
robotic system with hybrid tracking for MRI-guided prostate intervention. IEEE
Trans. Biomed. Eng., 58(11):3049–3060, 2011.
[51] A. Krieger, R. Susil, C. Menard, J. Coleman, G. Fichtinger, E. Atalar, and
L.L. Whitcomb. Design of novel MRI compatible manipulator for image guided
prostate interventions. IEEE Trans. on Biomed. Eng., 52(2):306–312, 2005.
[52] J.D. Kronz, C.H. Allan, A.A. Shaikh, and J.I. Epstein. Predicting cancer following a diagnosis of high grade prostatic intraepithelial neoplasia on needle
biopsy: data on men with more than one follow-up biopsy. Am. J. Surg. Pathol.,
25(8):1079–1085, 2001.
[53] V. Lagerburg, M. Moerland, J.J. Lagendijk, and Battermann J.J. Measurement
of prostate rotation during insertion of needles for brachytherapy. Radiother.
Oncol., 77(3):318–323, 2005.
[54] V. Lagerburg, M.A. Moerland, M. van Vulpen, and J.J.W. Lagendijk. A new
robotic needle insertion method to minimise attendant prostate motion. Radiother. Oncol., 80:7377, 2006.
BIBLIOGRAPHY
93
[55] J.H. Lynch and S. Loeb. The role of high-intensity focused ultrasound in
prostate cancer. Curr. Oncol. Rep., 9:222–225, 2007.
[56] Y. Mazaheri, L. Bokacheva, D.J. Kroon, O. Akin, H. Hricak, D. Chamudot,
S. Fine, and J.A. Koutcher. Semi-automatic deformable registration of prostate
mr images to pathological slices. J Magn. Reson. Imaging, 32(5):1149–1157,
2010.
[57] C. Menard, R.C. Susil, P. Choyke, G.S. Gustafson, W. Kammerer, H. Ning,
R.W. Miller, K.L. Ullman, N.S. Crouse, S. Smith, E. Lessard, J. Pouliot,
V. Wright, E. McVeigh, C.N. Coleman, and K. Camphausen. MRI-guided HDR
prostate brachytherapy in standard 1.5T scanner. Int. J. Radiation Oncology
Biol. Phys., 59(5):14141423, 2004.
[58] S. Misra, K. Macura, K. Ramesh, and A. Okamura. The importance of organ
geometry and boundary constraints for planning of medical interventions. Med.
Eng. Phys., 31(2):195–206, 2009.
[59] P. Mohan, H. Ho, J. Yuen, W.S. Ng, and W.S. Cheng. A 3D computer simulation to study the efficacy of transperineal versus transrectal biopsy of the
prostate. Int. J. CARS, 1:351360, 2007.
[60] J.A. Molloy and M.B. Williams. Treatment planning considerations and quality
assurance for CT-guided transischiorectal implantation of the prostate. Med.
Phys., 26(9):1943–1951, 1999.
BIBLIOGRAPHY
94
[61] P. Mozer, M. Baumann, and G. Chevreau et al. Accuracy evaluation of prostate
biopsies under endorectal 2D ultrasound. In 22nd Annual Meeting of the Engineering and Urology Society, Anaheim, CA, 2007.
[62] S. Natarajan, L.S. Marks, D.J.A. Margolis, J. Huang, M.L. Macairan, P. Lieu,
and A. Fenster. Clinical application of a 3D ultrasound-guided prostate biopsy
system. Urol. Oncol., 29(3):334–342, 2011.
[63] A. Nederveen, U. Van der Heide, H. Dehnad, R.J. Van Moorselaar, P. Hofman,
and J.J. Lagendijk. Measurements and clinical consequences of prostate motion during a radiotherapy fraction. Int. J. Radiation Oncology Biol. Phys.,
53(1):206–214, 2002.
[64] S.M. Noworolski, J.C. Crane, D.B. Vigneron, and J. Kurhanewicz. A clinical comparison of rigid and inflatable endorectal-coil probes for MRI and 3D
MR spectroscopic imaging (MRSI) of the prostate. J. Magn. Reson. Imaging,
27(5):1077–1082, 2008.
[65] S. Oguro, J. Tokuda, H. Elhawary, S. Haker, R. Kikinis, C. Tempany, and
N. Hata. MRI signal intensity based B-spline nonrigid registration for pre- and
intraoperative imaging during prostate brachytherapy. J. MRI., 30:1052–1058,
2009.
[66] L. Phee, D. Xiao, J. Yuen, C.F. Chan, H. Ho, C.H. Thng, C. Cheng, and W.S.
Ng. Ultrasound guided robotic system for transperineal biopsy of the prostate.
In IEEE International Conference on Robotics and Automation, pages 1315–
1320, Barcelona, Spain, 2005.
BIBLIOGRAPHY
95
[67] S. Pieper, M. Halle, and R. Kikinis. 3D slicer. In IEEE Int. Symposium on
Biomed. Imaging: From nano to macro, pages 632–635, 2004.
[68] P. Pinto, P.H. Chung, A.R. Rastinehad, A.A. Baccala Jr., J. Kruecker, C.J.
Benjamin, S. Xu, P. Yan, S. Kadoury, C. Chua, J.K. Locklin, B. Turkbey, J.H.
Shih, S.P. Gates, C. Buckner, G. Bratslavsky, W.M. Linehan, N.D. Glossop,
P.L. Choyke, and B.J. Wood. Magnetic resonance imaging/ultrasound fusion
guided prostate biopsy improves cancer detection following transrectal ultrasound biopsy and correlates with multiparametric magnetic resonance imaging.
J. Urol., 186(4):1281–1285, 2011.
[69] G. Ploussard, J.I. Epstein, R. Montironi, P.R. Carroll, M. Wirth, M.O. Grimm,
A.S. Bjartell, F. Montorsi, S.J. Freedland, A. Erbersdobler, and T.H. van der
Kwast. The contemporary concept of significant versus insignificant prostate
cancer. Euro. Urol., 60:291–303, 2011.
[70] T. Podder, D. Clark, J. Sherman, D. Fuller, E. Messing, D. Rubens, J. Strang,
R. Brasacchio, L. Liao, W.S. Ng, and Y. Yu. In vivo motion and force measurement of surgical needle intervention during prostate brachytherapy. Med Phys.,
33(8):2915–2922, 2006.
[71] T. Podder, J. Sherman, D. Rubens, E. Messing, J. Strang, W.S. Ng, and Y. Yu.
Methods for prostate stabilization during transperineal LDR brachytherapy.
Phys. Med. Biol., 53:15631579, 2008.
[72] A. Polo, C. Salembier, J. Venselaar, and P. Hoskin. Review of intraoperative
imaging and planning techniques in permanent seed prostate brachytherapy.
Radiother. Oncol., 94:12–23, 2010.
BIBLIOGRAPHY
96
[73] K.M. Pondman, J.J. Futterer, B. Ten Haken, L.J.S. Kool, J.A. Witjes, T. Hambrock, K.J. Macura, and J.O. Barentsz. MR-guided biopsy of the prostate: an
overview of techniques and a systematic review. Eur. Urol., 54(3):517–527,
2008.
[74] J. Pouliot, Y. Kim, E. Lessard, I-C. Hsu, D.B. Vigneron, and J. Kurhanewicz.
Inverse planning for HDR prostate brachytherapy used to boost dominant intraprostatic lesion defined by magnetic resonance spectroscopy imaging. Int. J.
Radiation Oncology Biol. Phys., 59(4):11961207, 2004.
[75] C. Reynier, J. Troccaz, P. Fourneret, A. Dusserre, C. Gay-Jeune, J.L. Descotes,
M. Bolla, and J.Y. Giraud. MRI/TRUS data fusion for prostate brachytherapy.
preliminary results. Med. Phys., 31(6):1568–1575, 2004.
[76] A. Sahgal and M. RoachIII. Permanent prostate seed brachytherapy: a current
perspective on the evolution of the technique and its application. Nat. Clin.
Pract. Urol., 4(12):658–670, 2007.
[77] S. Sang-Eun, N.B. Cho, I.I. Iordachita, P. Guion, G. Fichtinger, A. Kaushal,
K. Camphausen, and L.L. Whitcomb. Biopsy needle artifact localization in
MRI-guided robotic transrectal prostate intervention. IEEE Trans. Biomed.
Eng., 59(7):1902–1911, 2012.
[78] M. Shelley, T.J. Wilt, B. Coles, and M.D. Mason. Cryotherapy for localised
prostate cancer. Cochrane Database Syst. Rev., 3:CD005010, 2007.
[79] N. Shusharina and G. Sharp. Analytic regularization for landmark-based image
registration. Phys Med Biol, 57(6):1477–1498, 2012.
BIBLIOGRAPHY
97
[80] R. Siegel, E. Ward, O. Brawley, and A. Jemal. Cancer statistics, 2011 the
impact of eliminating socioeconomic and racial disparities on premature cancer
deaths. CA Cancer J Clin., 61:212–236, 2011.
[81] T. De Silva, A. Fenster, J. Bax, L. Gardi, C. Romagnoli, J. Samarabandu, and
A. Ward. 2D-3D rigid registration to compensate for prostate motion during
3d TRUS-guided biopsy. In SPIE, volume 8316, pages 83160O–1–6, 2012.
[82] T. De Silva, A. Fenster, J. Samarabandu, and A. Ward. Quantification of
prostate deformation due to needle insertion during TRUS-guided biopsy. In
MICCAI, volume 6363, pages 213–220. LNCS, 2010.
[83] A.K. Singh, J. Kruecker, S. Xu, N. Glossop, P. Guion, K. Ullman, P.L.
Choyke, and B.J. Wood. Initial clinical experience with real-time transrectal ultrasonography-magnetic resonance imaging fusion-guided prostate biopsy.
BJU International, 101(7):841–845, 2007.
[84] W.L. Smith, C. Lewis, G. Bauman, G. Rodrigues, D. D’Souza, R. Ash, D. Ho,
V. Venkatesan, Donal Downey, and A. Fenster. Prostate volume contouring: A
3D analysis of segmentation using 3DTRUS, CT, and MR. Int. J. Radiation
Oncology Biol. Phys., 67(4):1238–1247, 2007.
[85] American Cancer Society. Cancer facts and figures, 2012. Accessed on Jan 9,
2013.
[86] D.Y. Song, A.K. Jain, Z. Zhang, A. Deguet, Y. Le, E. Armour, E.C. Burdette,
and G. Fichtinger. Dynamic intraoperative dosimetry for prostate brachytherapy using a nonisocentric C-arm. Brachytherapy, 10(2):98–106, 2011.
BIBLIOGRAPHY
98
[87] N.N. Stone, J. Roy, S. Hong, Y.C. Lo, and R.G. Stock. Prostate gland motion
and deformation caused by needle placement during brachytherapy. Brachytherapy, 1(3):154–160, 2002.
[88] Y.S. Sung, H.J. Kwon, B.W. Park, G. Cho, C.K. Lee, K.S. Cho, and J.K.
Kim. Prostate cancer detection on dynamic contrast-enhanced MRI: computeraided diagnosis versus single perfusion parameter maps. AJR Am J Roentgenol,
197(5):1122–1129, 2011.
[89] R.C. Susil, C. Menard A. Krieger, J.A. Coleman, K. Camphausen, P. Choyke,
G. Fichtinger, L.L. Whitcomb, C.N. Coleman, and E. Atalar. Transrectal
prostate biopsy and fiducial marker placement in a standard 1.5T magnetic
resonance imaging scanner. J. Urol., 175:113–120, 2006.
[90] H. Tadayyon, A. Lasso, A. Kaushal, P. Guion, and G. Fichtinger. Target motion tracking in MRI-guided transrectal robotic prostate biopsy. IEEE Trans.
Biomed. Eng., 58(11):3135–3142, 2011.
[91] S. Takahashi, Y. Yamada, Y. Homma, S. Horie, Y. Hosaka, and T. Kitamura.
Power Doppler ultrasonography-directed prostate biopsy in men with elevated
serum PSA levels: an evaluation of the clinical utility and limitations. Urology,
60(2):248252, 2002.
[92] R. Taschereau, J. Pouliot, J. Roy, and D. Tremblay. Seed misplacement and
stabilizing needles in transperineal permanent prostate implants. Radiother.
Oncol., 55(1):5963, 2000.
BIBLIOGRAPHY
99
[93] G. Taverna, G. Morandi, M. Seveso, G. Giusti, A. Benetti, P. Colombo,
F. Minuti, F. Grizzi, and P. Graziotti. Colour Doppler and microbubble contrast agent ultrasonography do not improve cancer detection rate in transrectal
systematic prostate biopsy sampling. BJU International, 108(11):1723–1727,
2011.
[94] C. Tempany, S. Straus, N. Hata, and S. Haker. MR-guided prostate interventions. J. Magn. Reson. Imaging, 27:356367, 2008.
[95] M. Terris, E. Wallen, and T. Stamey. Comparison of mid-lobe versus lateral
systematic sextant biopsies in detection of prostate cancer. Urol. Int., 59:239242,
1997.
[96] M.K. Terris. Strategies for repeat prostate biopsies. Curr Urol Rep, 10(3):172–
178, 2009.
[97] C. Testa, R. Schiavina, and R. Lodi et al. Accuracy of MRI/MRSI-based transrectal ultrasound biopsy in peripheral and transition zones of the prostate
gland in patients with prior negative biopsy. NMR Biomed, 23(9):1017–1026,
2010.
[98] J. Tokuda, K. Tuncali, I. Iordachita, S.E. Song, A. Fedorov, S. Oguro, A. Lasso,
F.M. Fennessy, C.M. Tempany, and N. Hata. In-bore setup and software for
3T MRI-guided transperineal prostate biopsy. Phys. Med. Biol., 57:5823–5840,
2012.
[99] N.J. Tustison and J.C. Gee. N4ITK: Nicks N3 ITK implementation for MRI
bias field correction. Insight J., 2009. Accessed on Jan 9, 2013.
BIBLIOGRAPHY
100
[100] M.R. van den Bosch, M.R. Moman, M. van Vulpen, J.J. Battermann, E. Duiveman, L.J. van Schelven, H. de Leeuw, J.J.W. Lagendijk, and M.A. Moerland.
MRI-guided robotic system for transperineal prostate interventions: proof of
principle. Phys. Med. Biol., 55:N133N140, 2010.
[101] M. van Herk, A. Bruce, A.P.G. Kroes, T. Shouman, A. Touw, and J.V.
Lebesque. Quantification of organ motion during conformal radiotherapy of
the prostate by three dimensional image registration. Int. J. Radiat. Oncol.
Biol. Phys., 33:1311–1320, 1995.
[102] C. Vargas, A.I. Saito, W.C. Hsi, D. Indelicato, A. Falchook, Q. Zengm, K.Oliver,
S. Keole, and Jim Dempsey. Cine-magnetic resonance imaging assessment of
intrafraction motion for prostate cancer patients supine or prone with and without a rectal balloon. Am. J. Clin. Oncol., 33(1):11–16, 2010.
[103] P. De Visschere, W. Oosterlinck, G. De Meerleer, and G. Villeirs. Clinical and
imaging tools in the early diagnosis of prostate cancer, a review. JBR-BTR,
93(2):62–70, 2010.
[104] K. Wallner, J.C. Blasko, and M. Dattoli. Prostate brachytherapy made complicated. SmartMedicine Press, 2001.
[105] R.A. Weersink, A. Bogaards, M. Gertner, S.R.H. Davidson, K. Zhang,
G. Netchev, J .Trachtenberg, and B.C. Wilson. Techniques for delivery and
monitoring of TOOKAD (WST09)-mediated photodynamic therapy of the
prostate: clinical experience and practicalities.
79(3):211–222, 2005.
J Photochem Photobiol B,
BIBLIOGRAPHY
101
[106] A. Wefer, H. Hricak, D. Vigneron, F. Coakley, Y. Lu, J. Wefer, U. Mueller-Lisse,
P. Carroll, and J. Kurhanewicz. Sextant localization of prostate cancer: comparison of sextant biopsy, magnetic resonance imaging and magnetic resonance
spectroscopic imaging with step section histology. J. Urol., 163(2):400404, 2000.
[107] H. Welch, E. Fisher, D. Gottlieb, and M. Barry. Detection of prostate cancer
via biopsy in the medicareSEER population during the PSA era. J. Nat. Cancer
Inst., 99(18):1395–1400, 2007.
[108] H. Westendorp, C.J. Hoekstra, A. vant Riet, A.W. Minken, and J.J. Immerzeel.
Intraoperative adaptive brachytherapy of iodine-125 prostate implants guided
by C-arm cone-beam computed tomography-based dosimetry. Brachytherapy,
6(4):231–237, 2007.
[109] H. Xu, A. Lasso, P. Guion, A. Krieger, A. Kaushal, A.K. Singh, P.A. Pinto,
J. Coleman, R.L. Grubb III, J.B. Lattouf, C. Menard, L.L. Whitcomb, and
G. Fichtinger. Accuracy analysis in MRI-guided robotic prostate biopsy. Int J
Comput Assist Radiol Surg., 8(6):937–944, 2013.
[110] H. Xu, A. Lasso, S. Vikal, P. Guion, A. Krieger, A. Kaushal, L.L. Whitcomb,
and G. Fichtinger. Accuracy validation for MRI-guided robotic prostate biopsy.
In SPIE Medical Imaging, pages 762517–762518, 2010.
[111] H. Xu, A. Lasso, S. Vikal, P. Guion, A. Krieger, A. Kaushal, L.L. Whitcomb,
and G. Fichtinger. Clinical accuracy of robot-assisted prostate biopsy in closed
MRI scanner. In 3rd Hamlyn Symposium on Medical Robotics, pages 7–8, 2010.
BIBLIOGRAPHY
102
[112] H. Xu, A. Lasso, S. Vikal, P. Guion, A. Krieger, A. Kaushal, L.L. Whitcomb,
and G. Fichtinger. MRI-guided robotic prostate biopsy: a clinical accuracy
validation. In MICCAI, volume 6363, pages 383–391. LNCS, 2010.
[113] H. Xu, A. Lasso, S. Vikal, P. Guion, A. Krieger, A. Kaushal, L.L. Whitcomb,
and G. Fichtinger. MRI-guided transrectal robotic prostate biopsy validation.
In AAPM, volume 37, pages 3128–3129. Med. Phys., 2010.
[114] S. Xu, J. Kruecker, B. Turkbey, N. Glossop, A. Singh, P. Choyke, P. Pinto,
and B.J. Wood. Real-time MRI-TRUS fusion for guidance of targeted prostate
biopsies. Comput. Aided. Surg., 13(5):255–264, 2008.
[115] T.S. Yoo, M.J. Ackerman, W.E. Lorensen, W. Schroeder, V. Chalana, S. Aylward, D. Metaxes, and R. Whitaker. Engineering and algorithm design for
an image processing API: a technical report on ITKthe insight toolkit. Stud.
Health Technol. Inform., 85:586–592, 2002.
[116] S. Zangos, A. Melzer, K. Eichler, C. Sadighi, A. Thalhammer, B. Bodelle,
R. Wolf, T. Gruber-Rouh, D. Proschek, R. Hammerstingl, C. Muller, M.G.
Mack, and T.J. Vogl. MR-compatible assistance system for biopsy in a highfield-strength system: initial results in patients with suspicious prostate lesions.
Radiology, 259(3):903–910, 2011.