Download Three-dimensional cardiovascular image analysis

IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 21, NO. 9, SEPTEMBER 2002
1005
Three-Dimensional Cardiovascular Image Analysis
I. MOTIVATION FOR THE SPECIAL ISSUE
CCORDING to estimates from the World Health Organization (WHO), 17 million people around the globe die
of cardiovascular disease (CVD) each year [1]. In 1998, there
were 7.3 million deaths from heart attack and 5.1 million from
stroke. About 600 million people with high blood pressure are
at risk of heart attack, stroke, and cardiac failure. In 1999, CVD
contributed to one-third of global deaths. Low- and middle-income countries contributed to 78% of CVD deaths. By 2010, it
is estimated that CVD will be the leading cause of death in developing countries. Heart disease has no geographic, gender, or
socioeconomic boundaries [2].1
As CVD continues to be the world’s leading cause of death,
any effort to improve its early diagnosis and therapy will
have a highly beneficial impact on society. This issue of this
TRANSACTIONS is concerned with efforts from the medical
imaging and medical image analysis communities to facilitate
the diagnosis of CVD.
Nowadays, there exist many imaging techniques to perform
cardiovascular examinations [3], [4]. Ultrasound (US), singlephoton emission computed tomography (SPECT), computed
tomography (CT), and magnetic resonance imaging (MRI) are
definitely the most well-known and established techniques.
However, many recent advances in hardware, contrast agents,
and postprocessing algorithms are empowering these methods
by extending the frontiers of their applicability.
For instance, hardware improvements in MRI, CT, and US
nowadays allow faster imaging protocols, enabling real-time
dynamic three-dimensional (3-D) imaging of the heart. This
has been demonstrated with parallel MRI acquisition strategies
like SMASH [5], [6] and SENSE [7], [8], with multislice CT
imaging [9]–[12], and using piezoelectric two-dimensional
(2-D) arrays or 3-D probe tracking systems in US [13]–[15].
Among existing imaging techniques, MRI deserves particular
attention as the ideal “single-stop shop” for cardiac diagnosis
and is rapidly becoming a reality through improvements in MR
acquisition strategies. Currently, many clinical groups are pursuing a protocol that can produce a thorough assessment of myocardial structure, function, and perfusion; assessment of coronary artery anatomy and flow; and spectroscopic evaluation of
cardiac energetics in one MR patient examination within one
hour [3], [16]–[19].
Contrast agents are another area of research where recent
advances have made breakthrough contributions to improve
image quality and, hence, the diagnostic value of medical
images [20]. For instance, with the advent of blood-pool agents
in MR [21]–[23], longer scanning times are possible and therefore a full-body scan is now a feasible virtual field-of-view.
A
Digital Object Identifier 10.1109/TMI.2002.804442
1World
Health Organization, http://www.who.int/ncd/cvd.
Besides their ability to increase vascular contrast, blood pool
agents provide physiological information, including rate of
entry, rate of accumulation, and rate of elimination. In addition,
MR imaging with blood pool agents has proven to be of
significant value in the assessments of myocardial perfusion
and microvascular permeability. In ultrasonic imaging, contrast
agents have also been developed over the last two decades to
improve image quality [24]–[26]. The most common US contrast agent is in the form of gas-filled stabilized microbubbles
that interact with US and yield a great enhancement of the
backscatter level. There are also targeted contrast agents [27]
that expand the detectability and diagnosis of pathology from
a strict anatomical to biochemical basis.
The rapid progress in dynamic cardiovascular imaging techniques has lead to new challenges in the handling of the huge
amount of image data involved in comprehensive functional patient studies. For instance, MRI stress studies require the acquisition of several image sequences at various stress levels (usually a baseline acquisition, three to five stress levels, and a recovery acquisition). During each MRI acquisition, typically six
to eight cross-sectional slices are acquired through the heart
from base to apex, all with about 20 time phases over the cardiac
cycle, plus a number of long-axis cross-sections, adding up to
about 200 images per cardiac acquisition. In this situation, analyzing all these images becomes a formidable task even for the
most enthusiastic cardiologist, radiologist, or technician in order
to interpret the data and derive clinically useful information for
diagnosis or decision support for surgical or pharmacological
interventions. Additionally, it is needless to say that manual
analysis is subjective and therefore compromises the accuracy
and reproducibility of any quantitative measurements. Since the
number of highly trained personnel is limited in a specific clinical center, these lengthy and complicated procedures form the
bottleneck for a widespread usage of these modern technologies, despite their tremendous potential. The abovementioned
reasons have triggered a great demand for algorithms and tools
that simplify and further automate the accurate analysis of cardiac patient examinations: together with “single-stop shop” examinations, the radiologist/cardiologist is eagerly awaiting this
“single-button analysis,” which extracts relevant clinical information about the function of the heart and the status of the coronary arteries from these images, all within a matter of minutes.
Nevertheless, the importance of computerized techniques for
analysis of cardiovascular images does not only arise from a
need to manage huge amounts of information automatically. It
is natural that interpretation of images from modern imaging
techniques tends to follow historical patterns. For example,
global functional indexes like ejection fraction, left ventricular
volume, and mass or wall thickening have been used for several
decades, primarily since they are easy to compute or estimate
from simplified models of the heart derived from 2-D imaging
techniques like X-ray ventriculography. It is foreseeable that
in the near future, those parameters will remain essential to
0278-0062/02$17.00 © 2002 IEEE
Authorized licensed use limited to: UNIVERSITAT POMPEU FABRA. Downloaded on May 31,2010 at 10:36:12 UTC from IEEE Xplore. Restrictions apply.
1006
IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 21, NO. 9, SEPTEMBER 2002
guide diagnosis given the large amount of clinical evidence
accumulated on their basis over the years. However, the combination of new imaging techniques and powerful computerized
image-analysis tools will yield novel performance indexes and
quantitative and visual means to achieve more precise early
diagnosis and more intuitive interpretation of results. More
importantly, the advantage of these new indexes will stem from
the fact that they better exploit the localized dynamic information provided from modern 3-D spatiotemporal imaging
techniques and from the fusion of information from multiple
modalities. More and more often, we are witnessing that novel
image analysis techniques tend to incorporate knowledge from
other disciplines, like biomechanics and computational fluid
dynamics, to yield physically sound approaches where the
final quantitative analysis that one is pursuing is taken into
account from an early phase. Also, image-analysis techniques
are currently no longer only seen as a postprocessing step
providing no feedback to image acquisition. More and more
often, image-processing or analysis techniques are driving
the acquisition itself to prospectively compensate for image
artifacts. All of the previous is clear evidence that cross-fertilization with other fields is essential in medical image analysis.
The main motivation of this Special Issue is to highlight in a
monographic number some of the current efforts from the medical imaging and medical image-analysis communities to improve image acquisition and analysis in the cardiovascular domain and to move toward three-dimensional analysis of the morphology and function of the heart. As such, we believe that the
contributions from several experts in the field constitute a step
forward in improving healthcare by helping to combat cardiovascular diseases.
II. BACKGROUND STATISTICS
The first call for papers of the Special Issue on Three-Dimensional Cardiovascular Image Analysis was announced in
the IEEE TRANSACTIONS ON MEDICAL IMAGING June 2001
issue. It was also advertised at several important conferences
in the field such as Medical Image Computing and Computer
Assisted Interventions 2001, Information Processing in Medical Imaging 2001, and Computers in Cardiology 2001. A total
of 27 submissions were received in the period between October
29 and December 1, 2001. All papers have been reviewed by
at least three reviewers, and a large number of them by four
experts. The review process was completely electronical and
therefore anticipated the current electronic system used for regular issues by a couple of months. This, together with the excellent and timely response from the reviewers, helped to speed
up the review process enormously. By March 15, 2002, all authors were preliminarily notified on their papers. All papers
subject to minor or major revisions went back to the original
reviewers for a second round. The final notification was sent
to the authors by July 15, 2002. A total of 17 papers were
accepted; three manuscripts were withdrawn from the Special
Issue and went to the regular submission procedure given the
tight publication schedule for the Special Issue. Finally, seven
manuscripts were rejected. In addition to the 17 accepted manuscripts, we have also incorporated into the issue two papers
originally submitted as regular manuscripts before this Special
Issue was announced.
III. CONTENTS OF THE ISSUE
The articles in this issue contribute to several aspects in the
cardiovascular domain ranging from novel imaging techniques
over improvements in image acquisition and reconstruction to
quantitative image analysis and image compression. They also
cover several imaging modalities with special emphasis on magnetic resonance imaging, different types of 3-D echocardiography, and SPECT. Several papers are devoted to the evaluations
of new or existing algorithms.
In the area of image acquisition and reconstruction, the paper
by Tilg et al.presents a validation study of a novel imaging technique coined activation time imaging [29]. This method allows
for the noninvasive imaging of parameters of the cardiac electrical function and as such promises to be a valuable tool for
diagnosis, in particular of arrhythmias. The method proposed
in this paper combines dynamic geometric information of the
heart obtained from MR scans with electrical information derived from electrocardiogram surface mapping. The result is a
spatiotemporal distribution of electrical activity that can facilitate the identification of single-focal, multifocal, and distributed
activation patterns. Both the noninvasive nature of the procedure, which obviously provides an advantage over uncomfortable long-lasting catheter procedures, and the fact that it can extract useful information from single-beat data contribute to the
clinical value of such a new tool.
The paper by Frey et al. presents a method to carry out a quantitative assessment of task-based image quality in myocardial
perfusion SPECT studies. This method is based on a model called
the channelized Hotteling observer, which is fundamental in
widely accepted psychophysiological experiments. This model
can be used to assess the quality of a SPECT image with respect
to the detection of perfusion defects. Having a measure of image
quality paves the way to evaluate and optimize image-reconstruction techniques and to compare the added value of different
types of reconstruction-based compensation methods to correct
for attenuation, detector response, and scattering. Using this
method, the authors quantitatively demonstrate the improvement
in detectability introduced by simultaneous compensation of all
the previous factors over partial compensation techniques. The
results provide evidence that more detailed clinical and human
observer studies of reconstruction-based compensation methods
would be highly desirable.
Several papers deal with methods to automate the analysis
of cardiac function. Model-based techniques appear to be the
dominant contributions in this issue. A number of papers deal
with quantitative analysis of global functional parameters from
different modalities. Gerard et al.present a model-based techsegmentation of the left ventricle
nique to approach 3-D
from real-time 3-D echocardiography. This paper presents a fast
method to segment the left ventricle (LV), which is based on
the elastic deformation of a two-simplex LV surface using a Lagrangian framework. The main contribution of this paper is a
temporal deformation model of the heart that is obtained from
a population of MR tag sequences. This allows for a closer ini-
Authorized licensed use limited to: UNIVERSITAT POMPEU FABRA. Downloaded on May 31,2010 at 10:36:12 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 21, NO. 9, SEPTEMBER 2002
tialization of the LV mesh in each time frame of the 3-D sequence by a predictive model of the heart dynamics. The paper
by Mitchell et al. presents a 3-D extension of the popular active
appearance model initially introduced by Edwards et al. [28].
The method is applied to 3-D MR cardiac data as well as 2-D
US image sequences. The authors provide a thorough validation
in these two imaging modalities. They show good correlation
between manually obtained estimates of endo- and epicardial
volumes and left ventricular mass and those obtained by their
approach.
The paper by Frangi et al. describes a novel technique to
construct statistical shape models in 3-D and is specifically designed to work with multiple-part objects like the heart. The
paper describes a technique to build a shape atlas, which is subsequently used to define a dense set of landmarks. Landmarks of
the atlas are propagated to specific instances in a shape database
by volumetric nonrigid registration using two novel similarity
measures. This paper provides evidence regarding the convergence and uniqueness of the atlas construction procedure and
quantitatively demonstrates the ability of the method to identify
some anatomical landmarks.
The work by Corsi et al. focuses on a comprehensive validation of surface evolution and level-set techniques in cardiac
ultrasound. The authors carry out an in vivo and in vitro evaluation using real-time three-dimensional echocardiography and
show a very good agreement of volume measurements between
this technique and MRI-based manual measurements. Ye et al.
introduce a technique for quantitative analysis of global functional parameters using 3-D freehand echocardiography. The
novelty of this method stems from the fact that information from
multiple acoustic windows is fused together by simultaneously
constraining the deformation of a geometric model. The authors demonstrate the agreement between several quantitative
parameters computed with this technique with those obtained
by manual assessment in MR images. This paper indicates that
including information from multiple acoustic windows can improve the accuracy of volume estimates.
The paper of Sanchez-Ortiz et al. describes a fully automatic method for the extraction of left ventricular endocardial surfaces from sequences acquired with a 3-D rotational
probe echocardiographic system. The method combines multiscale fuzzy clustering with phase-based boundary detection.
The ejection fraction computed by this method is compared
with measurements derived from manual segmentation, from
a user-guided tracking algorithm, and from SPECT multigated
acquisition (MUGA). Again on ultrasound images, Wolf et al.
introduce a method named ROPES (Restricted Optimal Path
Exploring Segmentation). This technique allows for an interactive quantitative analysis of four-dimensional (4-D) images
from transesophageal echocardiography. The main contribution of this technique is that the required user interaction is
kept to a minimum and that the method is robust to variations
in parameter settings. This technique has been integrated in a
clinical workstation and evaluated on patient data. The paper
by Song et al.concerns the development of a Bayesian-based
method for 3-D surface extraction that avoids direct feature
extraction prior to surface fitting. Again, as in some of the
previous papers, the key idea is to construct a model of ex-
1007
pected shape from an exemplar database (in a similar manner
to a shape-space-based method) and to use this model to guide
the localization of the surface boundary. The method has been
tested on a database of 45 freehand ultrasound clinical cases
where the “ground truth” was manual delineation.
A number of papers specifically address the problems of motion, deformation, and strain-rate analysis. These papers can be
distinguished between contributions addressing the problem of
estimating cardiac dynamics and those related to recover respiratory motion.
This issue contains two papers devoted to analyzing cardiac
motion from MRI. The paper by Chen et al. describes a method
for analyzing MR tagging sequences. They cast the problem
of 3-D and 4-D tag detection and tracking as a maximum a
posteriori estimation problem based on a Markov random field
model. In this approach, temporal continuity in the image sequence and image features are used to recover a B-solid. Results
are reported on applying this technique in five pig studies. A
very different approach to motion analysis is the one presented
by Masood et al.. They introduce the method of virtual tagging
of phase contrast MR images. Virtual tags are numerically introduced by fitting an analytical model of the velocity field to
the MR velocity data. This smoothing process is constrained in
order to preserve mass and thus yield a physically consistent deformation field. This technique requires manual segmentation
of the myocardial border in the first frame only. Therefore, it
significantly reduces the postprocessing burden of conventional
MR tagging.
Several papers in this issue contribute to the field of strainrate imaging. The review paper by D’Hooge et al. surveys the
principles of ultrasound strain (rate) imaging, a new modality
for local quantification of myocardial deformation at high temporal resolution. This paper also surveys several clinical application areas of this technique, such as the analysis of ischemia, the quantification of regional myocardial deformation
in stress echocardiography, and the identification of stunned
myocardium. The paper by Selskog et al. also addresses the
problem of strain-rate imaging but from a different modality.
The authors present a method for computing and visualizing
the time-resolved three-dimensional strain-rate tensor from 3-D
phase contrast MRI. This technique extends prior attempts to
quantify strain rate from one-dimensional (1-D) (US) or 2-D
(MR) velocity data to 3-D. Time-resolved true 3-D strain-rate
analysis is an important step toward understanding tissue dynamics, since the strain rate is shown to be nonplanar and to vary
with time throughout the myocardium. Therefore, this technique
is argued to give better estimates of principal strain directions
than previous attempts in 1-D or 2-D.
Two papers were specifically concerned with understanding
respiratory motion. This topic is essential for improving image
acquisition by prospective compensation of respiratory artifacts. The paper by Manke et al. presents a study demonstrating the benefit of respiratory motion models to improve
coronary arteriography with MRI. One of the contributions of
that paper is to highlight the need for affine motion models,
which better explain the complex patient-specific respiratory motion patterns than current translational motion models.
Also, this paper shows how the recovery of the motion model
Authorized licensed use limited to: UNIVERSITAT POMPEU FABRA. Downloaded on May 31,2010 at 10:36:12 UTC from IEEE Xplore. Restrictions apply.
1008
IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 21, NO. 9, SEPTEMBER 2002
from a low-resolution image can be fed back into the imaging
system to achieve real-time motion compensation. This can
greatly reduce image artifacts that would be otherwise difficult to compensate for retrospectively. Along similar lines,
McLeish et al.also analyze respiratory motion patterns. This
study utilizes rigid and nonrigid image registration techniques
together with a subsequent statistical analysis of the calculated deformation fields. The authors confirm previous findings that predominant motion occurs in the craneocaudal direction. However, they also show that rotational motion can
also be considerable, particularly at the apex. The derived motion models could also be of interest in navigator techniques
for free-breathing scans.
Finally, there are three contributions to this issue that fall into
different areas. The review paper by Mäkelä et al. surveys the
field of image registration with a particular focus on cardiac
images. Registration in the cardiovascular domain has specific
challenges and can be regarded as more demanding than in other
areas like brain registration. Giving its elastically deforming nature, nonrigid registration techniques play a crucial role in this
field (see the work of McLeish et al. and Manke et al.in this
issue). Particular emphasis is given to the multimodal registration of MR images to SPECT or positron emission tomography
(PET) data and the serial registration of PET or SPECT scans
for thoracic and cardiac applications.
The paper by Cañero et al. tackles the problem of reconstructing the 3-D geometry of vessels from biplane X-ray angiography. This paper addresses several new issues: identification and modeling of sources of nonlinear distortions in angiographic projection, separation of fixed and orientation-dependent components, and a rigorous investigation into distortion correction methods. Moreover, the authors present a 3-D
vessel centerline reconstruction method from projection images.
The method does not require establishing of correspondence between homologous points, which is a tedious and subjective task
when performed manually. The method further corrects for distortions during the reconstruction process rather than requiring
projection images to be unwarped. The authors show that this
can drastically improve the reconstruction accuracy by incorporating a proper (un)distortion model.
Nowadays, the technological improvements in imaging techniques like MRI, CT, and US are facilitating the acquisition of
3-D dynamic studies that yield an enormous amount of data in
a short period. This fact, together with the increasing interest
in an all-digital approach to medical image transmission and
archival, increases the need for compression techniques to reduce the bandwidth and storage requirements of the modern
picture archiving and communication system (PACS) while preserving the diagnostic value of the images. The paper by Zeng
et al.turns the focus of the issue to this important topic of image
compression by describing an algorithm for the wavelet compression of 4-D arbitrarily sized echocardiographic data. While
many authors have explored the added gains in compression
made possible by exploiting correlations in the third and fourth
dimension, there are several novel aspects to this algorithm. In
particular, the modification of the zero-tree structure to accept
varying numbers of children and thus allow the compression of
arbitrarily sized data without padding is potentially very useful
in reducing computation time and memory requirements.
ALEJANDRO F. FRANGI, Guest Editor
University of Zaragoza
Division of Biomedical Engineering,
Aragon Institute of Engineering Research
Zaragoza, Spain
DANIEL RUECKERT, Guest Editor
Imperial College of Science, Technology and Medicine
Visual Information Processing, Department of Computing
London, U.K.
JAMES S.DUNCAN, Guest Editor
Yale University School of Medicine
Department of Diagnostic Radiology
New Haven, CT 06520 USA
ACKNOWLEDGMENT
The Guest Editors would like to thank the Editor-in-Chief
and the Associate Editor-in-Chief of the IEEE TRANSACTIONS
ON MEDICAL IMAGING for their support throughout the gestation of this issue. They would like also to express their gratitude to the 94 reviewers who participated in this issue; several of whom had to review more than one manuscript. Their
quick feedback enabled this issue to stay close to its ambitious
schedule. The cadre of reviewers was composed by people from
academia, clinical institutions, and industry who are actively involved in research and development in the cardiovascular field.
Our thanks go to A. Amini, S. Arridge, D. Atkinson, L. Axel, E.
Berry, L. Bidaut, D. Boughner, M. Brady, M. Breeuwer, V. Chalana, Y. Chen, P. Clarysse, T. Cootes, P. Cosman, Y. Coudière, J.
D’Hooge, B. Dawant, H. Delingette, T. Denney, J. Dijkstra, O.
Dössel, P. Edwards, M. Egmont-Petersen, A. Fenster, D. Firmin,
M. Fitzpatrick, D. Friboulet, A. Gee, O. Gerard, G. Germano,
J. Goutsias, J. Greenleaf, T. Gustavsson, M. Hastenteufel, D.
Hawkes, B. He, D. Hill, S. Horbelt, G. Huiskamp, M. Insana,
W. Kerwin, D. Kraitchman, A. Laine, B. Lelieveldt, Christian
Lorenz, Christine Lorenz, J. Lötjönen, I. Magnin, S. MakramEbeid, G. Malandain, A. Manduca, D. Manke, C. Maurer, A.
McCulloch, T. McInerney, G. Menegaz, J. Montagnat, E. Nagel,
B. Neyran, W. Niessen, A. Noble, W. O’Dell, M. O’Donnel,
C. Ozturk, X. Papademetris, J. Park, G. Penney, F. Pinciroli,
F. Rademakers, P. Radeva, R. Razavi, J. Ridgway, R. Rohling,
K. Rohr, F. Sachse, D. Sahn, G. Sanchez-Ortiz, J. Schnabel, F.
Sheehan, R. Shekhar, P. Shi, M. Sonka, L. Staib, M. Stegmann,
G. Stetten, M. Sühling, P. Summers, J. Suri, H. Torp, R. van der
Geest, A. Wahle, C. Xu, G-Z. Yang, and A. Young.
REFERENCES
[1] “2002 heart and stroke statistical update,” American Heart Association,
Dallas, TX, 2001.
[2] “European cardiovascular disease statistics,” British Heart Foundation,
2000.
[3] J. G. Goldin, O. Ratib, and D. R. Aberle, “Contemporary cardiac
imaging: An overview,” J Thorac. Imag., vol. 15, no. 4, pp. 218–29,
Oct. 2000.
Authorized licensed use limited to: UNIVERSITAT POMPEU FABRA. Downloaded on May 31,2010 at 10:36:12 UTC from IEEE Xplore. Restrictions apply.
IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 21, NO. 9, SEPTEMBER 2002
[4] SB. Reeder, Y. P. Du, J. A. Lima, and D. A. Bluemke, “Advanced cardiac
MR imaging of ischemic heart disease,” Radiographics, vol. 21, no. 4,
pp. 1047–74, July–Aug. 2001.
[5] D. K. Sodickson and W. J. Manning, “Simultaneous acquisition of spatial harmonics (SMASH): Fast imaging with radiofrequency coil arrays,” Magn. Reson. Med., vol. 38, pp. 591–603, 1997.
[6] D. K. Sodickson, “Spatial encoding using multiple RF coils: SMASH
imaging and parallel MRI,” in Methods in Biomedical Magnetic Resonance Imaging and Spectroscopy, I. Young, Ed. Chichester, U.K.:
Wiley, 2000, pp. 239–250.
[7] K. P. Pruessmann, M. Weiger, M. B. Scheidegger, and P. Boesiger,
“SENSE: Sensitivity encoding for fast MRI,” Magn. Reson. Med., vol.
42, pp. 952–962, 1999.
[8] M. Weiger, K. P. Pruessmann, and P. Boesiger, “Cardiac real-time
imaging using SENSE,” Magn. Reson. Med., vol. 43, pp. 177–184, 2000.
[9] K. Taguchi and H. Aradate, “Algorithm for image reconstruction in multislice helical CT,” Med. Phys., vol. 25, no. 4, pp. 550–61, Apr. 1998.
[10] H. Hu, “Multi-slice helical CT: Scan and reconstruction,” Med. Phys.,
vol. 26, no. 1, pp. 5–18, Jan. 1999.
[11] K. Klingenbeck-Regn, T. Flohr, B. Ohnesorge, J. Regn, and S. Schaller,
“Strategies for cardiac CT imaging,” Int. J. Cardiovasc. Imag., vol. 18,
no. 2, pp. 143–51, April 2002.
[12] K. Klingenbeck-Regn, S. Schaller, T. Flohr, B. Ohnesorge, K. Klingenbeck-Regn, and U. Baum, “Subsecond multi-slice computed
tomography: Basics and applications,” Eur. J. Radiol., vol. 31, no. 2,
pp. 110–24, Aug. 1999.
[13] W. Lees, “Ultrasound imaging in three and four dimensions,” Semin.
Ultrasound CT MR, vol. 22, no. 1, pp. 85–105, Feb. 2001.
[14] A. Fenster, D. B. Downey, and H. N. Cardinal, “Three-dimensional ultrasound imaging,” Phys. Med. Biol., vol. 46, no. 5, pp. R67–99, May 2001.
[15] A. Lange, P. Palka, D. J. Burstow, and M. J. Godman, “Three-dimensional
echocardiography: Historical development and current applications,” J.
Amer. Soc. Echocardiogr., vol. 14, no. 5, pp. 403–12, May 2001.
[16] J. Boxerman, TJ. Mosher, ER. McVeigh, E. Atalar, J. A. Lima, and D. A.
Bluemke, “Advanced MR imaging techniques for evaluation of the heart
and great vessels,” Radiographics, vol. 18, no. 3, pp. 543–64, May–June
1998.
1009
[17] C. Kramer, “Integrated approach to ischemic heart disease,” The
One-Stop Shop: Cardiol. Clin., vol. 16, no. 2, pp. 267–76, May
1998.
[18] E. Nagel and E. Fleck, “Magnetic resonance imaging (MRI) in cardiac
diagnostics,” Medica Mundi, vol. 45, no. 1, pp. 23–30, 2001.
[19] A. Bornstedt, E. Nagel, S. Schalla, B. Schnackenburg, C. Klein, and E.
Fleck, “Multi-slice dynamic imaging: Complete functional cardiac MR
examination within 15 seconds,” J. Magn. Reson. Imag., vol. 14, no. 3,
pp. 300–5, Sept. 2001.
[20] M. Marmion and E. Deutsch, “Tracers and contrast agents in cardiovascular imaging: Present and future,” Q. J. Nucl. Med., vol. 40, no. 1, pp.
121–31, Mar. 1996.
[21] LJ. Kroft and A. de Roos, “Blood pool contrast agents for cardiovascular
MR imaging,” J. Magn. Reson. Imag., vol. 10, no. 3, pp. 395–403, Sept.
1999.
[22] C. H. Lorenz and L. O. Johansson, “Contrast-enhanced coronary MRA,”
J. Magn. Reson. Imag., vol. 10, no. 5, pp. 703–8, Nov. 1999.
[23] M. Saeed, M. F. Wendland, and CB. Higgins, “Blood pool MR contrast
agents for cardiovascular imaging,” J. Magn. Reson. Imag., vol. 12, no.
6, pp. 890–8, Dec. 2000.
[24] J. Ophir and K. J. Parker, “Contrast agents in diagnostic ultrasound,”
Ultrasound Med. Biol., vol. 15, pp. 319–333, 1989.
[25] B. B. Goldberg, J. B. Liu, and F. Forsberg, “Ultrasound contrast agents:
A review,” Ultrasound Med. Biol., vol. 20, pp. 319–333, 1994.
[26] C. X. Deng and F. L. Lizzi, “A review of physical phenomena associated
with ultrasonic contrast agents and illustrative clinical applications,” Ultrasound Med. Biol., vol. 28, no. 3, pp. 277–86, Mar. 2002.
[27] G. M. Lanza and S. A. Wickline, “Targeted ultrasonic contrast agents
for molecular imaging and therapy,” Prog. Cardiovasc. Dis., vol. 44, no.
1, pp. 13–31, July–Aug. 2001.
[28] G. J. Edwards, T. F. Cootes, and C. J. Taylor, “Active appearance
models,” IEEE Trans. Pattern Anal. Machine Intell., vol. 23, no. 6, pp.
681–685, 2001.
[29] R. Modre, B. Tilg, G. Fischer, and P. Wach, “Noninvasive myocardial activation time imaging: a novel inverse algorithm applied to clinical ECG
mapping data.,” IEEE Trans. Biomed. Eng., vol. 49, pp. 1153–1161, Oct.
2002.
Alejandro F. Frangi (S’95–A’01) received the M.Sc. degree in Telecommunication Engineering
from the Universitat Politècnica de Catalunya, Barcelona, Spain, in 1996 and the Ph.D. degree
from the Image Sciences Institute, University Medical Center Utrecht, the Netherlands, in 2001.
He subsequently did research at Universitat Politècnica de Catalunya on electrical impedance
tomography for image reconstruction and noise characterization. His doctoral work was on
model-based cardiovascular image analysis. He is now Assistant Professor at the University
of Zaragoza, Spain, and is affiliated with the Division of Biomedical Engineering, Aragon
Institute of Engineering Research (I3A), a multidisciplinary research institute of the University
of Zaragoza. He currently leads the Visual Information Technologies Lab. His main research
interests are in computer vision and medical image analysis with particular emphasis in modeland registration-based techniques and statistical methods.
Dr. Frangi has recently been awarded the Ramón y Cajal Research Fellowship, a national
program of the Spanish Ministry of Science and Technology to promote outstanding young
investigators.
Daniel Rueckert received the M.Sc. degree from the Technical University, Berlin, Germany, in
1993 and the Ph.D. degree from Imperial College of Science, Technology and Medicine, London,
U.K., in 1997, both in computer science.
While pursuing the Ph.D. degree, he worked on the segmentation, motion tracking, and analysis of cardiovascular MR images. From 1997 to 1999, he was a Postdoctoral Research Fellow
in the Computational Imaging Science Group, Guy’s Hospital, developing algorithms for the
nonrigid registration of dynamic MR mammography images and for the compensation and quantification of tissue deformation between pre- and postoperative MR images. He is now a Lecturer
(Assistant Professor) with the Visual Information Processing group, Department of Computing,
Imperial College. His research interests include nonrigid registration of cardiac as well as neurological images and model-based image segmentation and analysis techniques.
Authorized licensed use limited to: UNIVERSITAT POMPEU FABRA. Downloaded on May 31,2010 at 10:36:12 UTC from IEEE Xplore. Restrictions apply.
1010
IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 21, NO. 9, SEPTEMBER 2002
James S. Duncan (M’84–SM’92–F’01) received the B.S.E.E. degree from Lafayette College,
Easton, PA, in 1973, the M.S. degree in engineering from the University of California, Los Angeles, in 1975, and the Ph.D. degree in electrical engineering from the University of Southern
California, Los Angeles, in 1982.
In 1973, he joined the Electro-Optical and Data Systems Group, Hughes Aircraft Company,
and participated in research and development projects related to signal and image processing for
forward-looking infrared imaging systems. During this time, he held Hughes’ Masters, Engineer,
and Doctoral Fellowships. In 1983, he joined the Faculty of Yale University, New Haven, CT,
where he currently is a Professor of diagnostic radiology and electrical engineering, Director
of the Image Processing and Analysis Group within Diagnostic Radiology, and Chair of Yale’s
Program in Biomedical Engineering. His research and teaching efforts have been in the areas of
image processing, computer vision, and medical imaging. His current specific research interests
include the segmentation of deformable objects from both 2-D and 3-D data, the tracking of nonrigid object motion from 2-D and 3-D data, the use of geometrical and physical models to constrain the recovery of information
from images, and the integration of processing modules in vision systems, all with a special interest in using these approaches for
medical image analysis.
Dr. Duncan is a Fellow of the American Institute for Medical and Biological Engineering. He is a member of Eta Kappa Nu and
Sigma Xi. He is on the editorial board of the Journal of Mathematical Imaging and Vision. He is an Associate Editor for the IEEE
TRANSACTIONS ON MEDICAL IMAGING and a Coeditor of Medical Image Analysis. He was a Fulbright Research Scholar at the
Universities of Amsterdam and Utrecht in the Netherlands during part of the 1993–1994 academic year. In June 1997, he chaired
the international conference on Information Processing in Medical Imaging, Poultney, VT.
Authorized licensed use limited to: UNIVERSITAT POMPEU FABRA. Downloaded on May 31,2010 at 10:36:12 UTC from IEEE Xplore. Restrictions apply.