Download Informatics in Radiology: Sliding-Thin

Note: This copy is for your personal non-commercial use only. To order presentation-ready
copies for distribution to your colleagues or clients, contact us at www.rsna.org/rsnarights.
INFORMATICS
317
Informatics in Radiology
Sliding-Thin-Slab Averaging for
Improved Depiction of Low-Contrast
Lesions with Radiation Dose Savings at
Thin-Section CT1
Christian von Falck, MD • Michael Galanski, MD, PhD • Hoen-oh Shin, MD
TEACHING
POINTS
See last page
Current multidetector computed tomography (CT) scanners allow
volumetric data acquisition with thin-section collimations and overlapping section reconstructions. The resultant nearly isotropic data sets
help minimize partial-volume averaging effects and are ideal for twoand three-dimensional postprocessing and software-assisted lesion detection and quantification. However, the section thickness, image noise,
and radiation dose are closely related, and when one parameter must
be altered to suit the clinical setting, the others may be affected. When
the clinical purpose demands both high spatial resolution and low image noise (eg, for the detection of hypoattenuating lesions in organs
such as the kidneys and liver), the necessary trade-off—an increase in
the radiation dose to the patient—may be unacceptable. The application
of a sliding-thin-slab averaging algorithm during image postprocessing and review helps overcome this limitation by reconstructing thicker
sections with lower noise levels from thin-section data obtained with
dose-saving protocols. In principle, a high noise level is acceptable in the
initial reconstruction of the CT volume data set. During image review
at the workstation, the section thickness can be interactively increased
to minimize image noise and improve lesion detectability. The combination of thin-section scanning with thick-section display allows routine
volumetric imaging without a general increase in radiation dose or a
reduction in the detectability of low-contrast lesions. Supplemental material available at http://radiographics.rsna.org/lookup/suppl/doi:10.1148
/rg.302096007/-/DC1.
©
RSNA, 2010 • radiographics.rsna.org
Abbreviation: SNR = signal-to-noise ratio
RadioGraphics 2010; 30:317–326 • Published online 10.1148/rg.302096007 • Content Codes:
From the Department of Radiology, Hannover Medical School, Carl-Neuberg-Strasse 1, 30625 Hannover, Germany. Received May 1, 2009; revision
requested August 3 and received September 9; accepted September 17. All authors have no financial relationships to disclose. Address correspondence
to C.V.F. (e-mail: [email protected]).
1
©
RSNA, 2010
318 March-April 2010
Introduction
Multidetector computed tomography (CT)
scanners with 64 or more detector rows enable
the routine use of submillimetric collimation for
image acquisition and the reconstruction of highresolution isotropic image data sets. The resultant
data sets help minimize through-plane partialvolume averaging effects and are optimally suited
for two- and three-dimensional postprocessing
with techniques such as multiplanar reformatting
and interactive volume rendering, as well as for
software-assisted lesion detection and quantification (1–5). However, image noise is inversely
related to section thickness; thus, a decrease in
the acquired section thickness leads to increased
image noise, which can be offset only by altering
the image acquisition protocol in ways that result
in an increased radiation dose to the patient. On
one hand, radiation dose control in accordance
with the ALARA (as low as reasonably achievable) principle demands careful adjustment of the
acquisition protocol to allow only the exposure
necessary to answer the specific clinical question.
On the other hand, the clear depiction of lowcontrast objects is essential to many multidetector
CT applications (eg, abdominal imaging evaluations), and diagnostic fidelity may be substantially
degraded by increased image noise (6,7).
We recommend that thin-section scanning be
combined with sliding-thin-slab averaging during
data postprocessing and image display, to gain
both high through-plane spatial resolution with
minimal partial-volume averaging effects and
excellent depiction of low-contrast lesions. In
the initial stage of data postprocessing, a secondary “raw” data set is reconstructed from a CT
volume data set that consists of overlapping sections with a thickness of 1 mm or less, acquired
at standard radiation dose levels comparable
to those for scanning with thicker sections (eg,
5 mm) (8). A relatively high degree of noise is
acceptable on the images from this initial reconstruction. On many commercially available workstations, the section thickness can be interactively
modified in real time by using the sliding-thinslab averaging algorithm. This approach allows
greater flexibility in multidetector CT scanning,
with the scanning protocol defining only the
radiographics.rsna.org
lower limit for section thickness. Thin-slab images
can easily be generated in any plane necessary for
the assessment of parenchymal organs, without
sacrificing through-plane resolution.
The article reviews the interdependence of radiation dose, image noise, and lesion detectability
and highlights the difficulty of objectively measuring the detectability of low-contrast objects on CT
scans. Various approaches for improving low-contrast lesion detectability are summarized, and the
sliding-thin-slab technique is described in detail.
Relations between Image Noise,
Image Quality, and Patient Dose
Image noise, patient dose, and image quality are
closely related. According to basic signal processing theory, noise generally impairs the detection
of signals (9). On multidetector CT scans, image
noise impairs the depiction of lesions in an organ
or tissue. Much higher image noise levels can be
tolerated in the depiction of organs with physiologic high contrast, such as that between air and
soft tissue in the lung, than in that of organs with
low contrast, such as the liver (Fig 1) (10,11).
A simple but useful descriptor of CT image
quality with respect to noise is the signal-to-noise
ratio (SNR), which is the ratio of the signal (the
mean attenuation value measured in Hounsfield
units in a region of interest within a lesion) to
noise (the standard deviation of the attenuation
values in a region in the background) (9). The
relationship between the SNR and the applied
dose (D) can be defined as follows:
SNR ∝ √D ⇔ D ∝ SNR2.
(1)
A more sophisticated formula by Brooks and
DiChiro (12) characterizes the interdependence
of patient dose, scanning parameters, and image
quality as
D∝
B
,
σ2 • a2 • b • h
(2)
where D is the radiation dose to the patient, B is
the object attenuation factor equivalent to
exp(µ · d), µ is the mean attenuation coefficient of
the object, d is the object thickness, σ is the variance of the CT numbers (in Hounsfield units),
a is the distance between two sampling points
measured at the isocenter of the scanner, b is the
effective beam thickness (detector collimation),
and h is the section thickness.
RG • Volume 30 Number 2
von Falck et al 319
Figure 1. (a, b) Axial multidetector CT images of the lungs show no substantial improvement in the
detectability of high-contrast lesions when image noise is reduced by increasing the reconstructed section thickness from 1 mm (a) to 5 mm (b). (c, d) Axial multidetector CT images at the level of the liver
show a marked improvement in the detectability of low-contrast lesions (arrows) when image noise is
reduced by increasing the reconstructed section thickness from 1 mm (c) to 5 mm (d). Not only are the
large lesions better delineated in d, but three smaller lesions (bottom arrows) are visible.
As mentioned earlier, image noise degrades
the detectability of low-contrast lesions at CT
(Fig 1). In theory, a variable K can be defined
that represents the threshold attenuation value at
which a lesion on a CT image becomes detectable, at a defined level of statistical probability,
against the surrounding background attenuation. Obviously, the quantity of image noise
alone, whether it is calculated as the variance of
attenuation values or as the SNR, is not a sufficient descriptor of lesion detectability. Other
noise characteristics, such as the image texture
or “granularity,” must be taken into account
(10,11). A mathematically precise description of
the magnitude of image noise can be achieved
by calculating the noise power spectrum, which
is determined by the spatial frequency of noise
on the CT image (13,14). The interdependence
of object size, noise characteristics, and lesion
detectability is illustrated in Figure 2.
320 March-April 2010
radiographics.rsna.org
Figure 2. Diagrams show the effect of noise characteristics on lesion detectability. The lesion
(black dot) is much more visible against a background of high-spatial-frequency noise (a) than
against a background of low-spatial-frequency noise (b), although the noise magnitude (calculated as the standard deviation of the background gray-scale values) is similar.
Since most parenchymal lesions seen in the
clinical setting are roughly spherical, we can safely base our calculations on the mathematical approximation of a spherical object. The detectability threshold K of a spherical lesion is dependent
on the lesion diameter d and inversely related to
the SNR and the square root of the applied dose
D, as shown in the following equation (9):
Kd ∝
1
1
.
∝
√D SNR
(3)
However, the dependence of the detectability
threshold K on lesion size d cannot be readily
calculated, because noise on CT images is not
mathematically uniform and does not follow a
defined statistical distribution (13,14). As a consequence, lesion detectability can be measured
only by comparing the results of multiple human
readings, performing various statistical analyses,
or creating experimental models of the human visual system, as described in the following section.
Measuring the Detectability of Low-Contrast Objects
It is widely accepted that the detectability of highcontrast objects such as a wire phantom at CT can
be quantified by using objective measures such as
the modulation transfer function. However, there
is less agreement about approaches for objectively
determining the detectability of low-contrast
objects at CT (9–11). For purposes of scanner
quality control and scientific studies, analyses of
receiver operating characteristic curves in reader
studies are the method of choice. Although the
accuracy of results is limited by the intra- and
interobserver variability of human readers, this approach is still the reference standard for assessing
low-contrast performance and may be appropriate
for clinical and experimental studies with a limited
number of cases. However, in larger studies (with
hundreds or thousands of images), classic readerbased analyses are not practical (10,11).
Alternative approaches have been reported but
not yet widely validated. One approach, proposed
by the American Society for Testing and Materials (15) and a CT scanner manufacturer (16), is
based on the use of statistical analysis of background noise levels within a phantom to quantify
the effects of various postprocessing techniques on
the detectability of low-contrast objects (10,11).
This method allows the analysis of a larger
number of data sets and eliminates the intra- and
interobserver variability found in studies based on
human readings. More sophisticated models of
the human visual system also have been used to
estimate image quality and lesion detectability, but
clinical validation is still lacking (17,18).
Optimizing the Detectability of Low-Contrast Objects
Various means are available for improving image quality and low-contrast lesion detectability
by reducing noise at thin-section multidetector
CT: For example, during image acquisition, an
increase in radiation dose to the patient can help
improve image quality by reducing quantum
RG • Volume 30 Number 2
von Falck et al 321
Figure 3. Graphs show the theoretical relationship of radiation exposure and section thickness in constant (a) and variable (b) noise conditions. When image noise is kept constant and
section thickness is reduced, the radiation dose to the patient increases in inverse proportion
to the reduction of section thickness, as shown in a. When image noise is allowed to increase,
the radiation dose ceases to be dependent on section thickness, as shown in b. Effects of scanner and beam geometry such as overbeaming and overranging are not reflected in the curves.
noise and increasing the SNR; during image data
reconstruction, a “soft” reconstruction kernel
may be used and various data filters applied; and
during image display and review, the window and
level settings may be adjusted and sliding-thinslab averaging may be applied to optimize the
depiction of low-contrast objects.
Changes made in the acquisition protocol to
improve image quality by reducing the magnitude
of noise and increasing the SNR result in an additional radiation burden to the patient, as demonstrated by Equation 1. An increase in the SNR
by a factor of 1.4 results in doubling of the radiation dose. The acceptability of a dose increase
depends on the clinical situation, but routine use
of high radiation dose levels at thin-section multidetector CT is not recommended (19).
By contrast, the choice of an appropriate “soft”
reconstruction kernel can substantially reduce
image noise and improve the detectability of lowcontrast lesions without affecting the radiation
dose to the patient (9,20). Although the reconstruction algorithm does not directly influence
the patient dose, the selection of the appropriate
kernel can help improve diagnostic accuracy at
multidetector CT performed with dose-saving
protocols tailored to the specific clinical situation. Because the properties of reconstruction
algorithms are not standardized and vary greatly
among vendors and scanner types, no general
recommendations can be made with regard to
optimal settings (20). In principle, the choice of a
specific reconstruction algorithm involves a tradeoff between the desired spatial resolution and the
acceptable quantity of image noise.
Window width and level settings at the workstation are closely related to the image noise level and
lesion detectability. According to various observers, the window width is inversely proportional to
the noise magnitude (21,22). Soft copy reading is
now standard in many radiology departments, and
all multimodality and picture archiving and communication system workstations allow interactive
adjustment of the window width and level settings
for optimal display and interpretation of a particular image data set. For hard copy–based readings
of liver studies, the use of an additional window is
recommended by some authors (23,24).
Although the choice of reconstruction algorithm is limited to the options available on the
CT scanner, additional data filters can be readily applied to reduce image noise and improve
low-contrast lesion detectability. However, use of
the wrong filter may degrade image quality and
reduce diagnostic fidelity (25). There is as yet no
standardized approach for such data filtering, and
detailed statistical analyses are needed to prove
diagnostic effectiveness.
In principle, the section thickness does not
affect the radiation burden to the patient as long
as volume coverage and all other scanning parameters remain unchanged. However, when section
thickness is halved and image quality (ie, image
noise) is to be kept constant, the patient dose must
double (Eq 2; Fig 3). By contrast, the acquisition of thicker sections can lead to a significant
improvement in image quality (ie, an increase in
SNR) without an increase in patient dose. The
322 March-April 2010
radiographics.rsna.org
Figures 4, 5. (4) Three-dimensional drawing shows the phantom (QRM, Moehrendorf, Germany) used
to obtain the images in Figures 5, 7, and 9. The phantom contains multiple solid spheres with diameters
of 3, 4, 5, 6, and 8 mm, which are embedded in a semisolid medium. Attenuation is calibrated to 15 HU
for the spheres and 35 HU for the background, with a resultant object contrast of 20 HU to simulate the
appearance of hypoattenuating lesions. The spheres are positioned at intervals throughout the x-y plane
and along the z-axis to allow the evaluation of low-contrast performance not only in the primary (axial)
reconstruction plane but also in sagittal and coronal planes. (5) Multidetector CT image series obtained in
the low-contrast phantom demonstrates the interdependence of section thickness and image quality. Noise
increases markedly with decreasing reconstructed section thickness from 5 to 3 to 1 mm (left to right)
when the radiation dose is kept constant (top row), whereas the SNR remains constant when the dose is
increased (bottom row). As symbolized by the black line below the bottom row of images, the dose used
with a section thickness of 1 mm is five times that used with a section thickness of 5 mm.
major drawback of increasing the section thickness
is a resultant increase in through-plane partialvolume averaging effects, which degrade multiplanar reformatted images and three-dimensional
volume-rendered images. Partial-volume averaging
effects also increase the likelihood that small lesions will be missed (9–11) (Figs 4, 5).
Overranging and overbeaming, two other
effects of multidetector CT scanning, also may
undermine dose efficiency at CT (26,27). Overranging refers to additional gantry rotations that
are automatically performed by the scanner to
acquire enough data for image reconstruction
when the section thickness and pitch selected for
reconstruction exceed the user-planned extension
of scanning along the z-axis (26). The number of
additional rotations varies, depending on the scan
parameters; it generally increases with increasing collimation, increasing section thickness in
the primary reconstruction, and increasing pitch.
The term overbeaming describes another impairment of CT scanner geometry, with a resultant
penumbra effect on images. It is well known that
four-channel multidetector CT scanners suf-
fer from inferior dose efficiency when used for
thin-section scanning, with a resultant increased
radiation dose as much as twice that incurred by
scanning with thicker sections. However, when
thin-section scanning is performed on multidetector CT scanners with 16 or more detector
rows, the overbeaming effect is much diminished.
The choice of section thickness at thin-section
CT scanning involves a trade-off between the
benefits of nearly isotropic voxels on one hand
and the disadvantages of image noise and radiation dose on the other. However, the use of a
sliding-thin-slab averaging algorithm at image reconstruction allows the advantage of thin-section
scanning, namely the reduction of through-plane
partial-volume averaging effects, to be retained,
while the detectability of low-contrast objects is
improved by the retrospective generation of thicker sections. It has been reported that the rate of
detection of low-contrast objects can be increased
by a factor of 1.1–1.7 over that based on standard
axial reconstructions, an improvement comparable to that achieved with an increase in radiation
dose by a factor of 1.2–2.9 (11). The technique
of sliding-thin-slab averaging is described in more
detail in the next section.
RG • Volume 30 Number 2
von Falck et al 323
Figure 6. Diagrams show the principles underlying the application of the sliding-thin-slab averaging algorithm.
A volumetric CT data set may be visualized as a stack of several hundred to more than one thousand axial image
sections obtained at regular intervals along the z-axis (a). The sliding-thin-slab averaging algorithm is used to reconstruct a subvolume (slab) that is located between two parallel clipping planes perpendicular to the arbitrarily
chosen viewing direction; the clipping planes are selected by the reader to define the location and thickness of the
slab, which should be greater than the acquired section thickness. All voxels that are located along the viewing axis
and within the subvolume are averaged and displayed as a single “thick” section (average intensity projection) so as
to minimize image noise (b). During interactive (cine) image review, the subvolume can be moved along the viewing axis in increments equivalent to the reconstruction interval to allow visualization of the entire lesion. (Ideally, the
increment should be 30%–50% of the acquired section thickness, when overlapping reconstruction is performed.)
With this method, image noise can be reduced without a corresponding increase in partial-volume averaging effects.
Sliding-Thin-Slab
Averaging Technique
The sliding-thin-slab averaging technique is
applied to a viewer-selected subvolume of the
acquired CT data set. The thickness of this
subvolume or slab is determined by the distance
between two parallel clipping planes that are perpendicular to the arbitrarily chosen viewing direction. Various rendering methods, such as average,
maximum, and minimum intensity projections
or volume rendering techniques, may be applied
to the selected slab (28). When the sliding-thinslab averaging algorithm is used, the trade-off between spatial resolution and image noise can be
interactively optimized in response to the varying
demands that arise during the reading process
(11) (Fig 6). A major advantage of this approach
is that the slab can be moved along the viewing
direction in increments (steps) of 1 mm or less to
allow depiction of the entirety or the major part
of a lesion so as to minimize partial-volume averaging effects. In contrast, the primary acquisition
of thicker sections and the use of larger recon-
struction intervals automatically lead to a larger
step interval and increased partial-volume averaging effects (Movie 1 [online]).
The usefulness of the sliding-thin-slab averaging algorithm is not restricted to the reconstruction of axial images but is equally relevant to that
of multiplanar and oblique views. Indeed, by
retaining the high through-plane spatial resolution of thin-section acquisitions while decreasing image noise and increasing the SNR, this
technique enables substantial improvement in
the detectability of low-contrast lesions on both
axial and reformatted images. The approach provides obvious clinical advantages (Figs 7, 8). Yet
another factor influences the quality of images
reconstructed with the sliding-thin-slab averaging
algorithm: In the primary (axial) reconstruction
of raw image data, image noise is correlated along
the z-axis.
324 March-April 2010
radiographics.rsna.org
Figure 7. Benefits of using the sliding-thin-slab averaging algorithm for multiplanar reconstruction of thin-section CT data. (a, b) Axial (top) and coronal (bottom) images reconstructed from a volume data set obtained in the low-contrast phantom show greatly reduced
SNR and severely deteriorated visibility of low-contrast objects with a reconstructed section
thickness of 1 mm (a) in comparison with those obtained with a reconstructed section thickness of 5 mm (b). However, relatively poor through-plane resolution results in unclear
delineation of the low-contrast objects in the coronal image in b. (c) Axial (top) and coronal
(bottom) images obtained by applying the sliding-thin-slab algorithm (slab thickness, 5 mm)
to the reconstructed data set used in a show high through-plane resolution with a higher SNR
and better visibility of low-contrast objects, especially in the coronal view.
Therefore, noise is even more effectively suppressed on sagittal and coronal images than on
axial images reconstructed with the sliding-thinslab averaging algorithm (Figs 9, 10).
Although the sliding-thin-slab technique has
been implemented on many commercially available multimodality and picture archiving and
communication workstations, there are hardly
any objective data in the literature to indicate an
effect on reader performance in the detection of
low-contrast lesions. Previous studies were focused mainly on the use of this technique to obtain maximum or minimum intensity projections
for pulmonary evaluations (29,30). Lee et al, in
two separate studies (31, 32), showed that the
sliding-thin-slab display mode enhances reader
confidence when compared with the stack mode
in the diagnosis of acute appendicitis (31) and
that the summation of thin sections is equivalent
to the primary reconstruction of thicker sections
(32). More recently, it was demonstrated that
the optimal slab thickness depends on the size
of the lesion to be detected and that the slab
thickness should be interactively varied during the reading process at the workstation (11)
(Movie 2 [online]). Optimal lesion detectability
is obtained with a slab thickness that is one-half
the lesion diameter.
The fact that the sliding-thin-slab averaging
algorithm is not yet available on all commercially
available workstations may be a deterrent to the
routine use of this postprocessing technique in
the interpretation of multidetector CT studies.
However, the CT workflow can be optimized by
using the primary reconstructed thin-section data
set to generate thin-slab multiplanar reformatted
images in standard orientations (adapted for optimal evaluation of the specific region) directly at
the scanner console. The necessary steps may be
completed by the technician in accordance with
predefined protocols or may be incorporated in
automated postprocessing programs on scanners
of recent generations. Although this alternative
approach does not permit interactive optimization of section thickness by the reader, it is a reasonable compromise when the sliding-thin-slab
averaging algorithm is not available at the picture
archiving and communication workstation.
Conclusions
Sliding-thin-slab averaging allows the reconstruction of high-resolution multiplanar images from
thin-section CT data sets without necessitating
an increase in radiation dose to preserve visibility
of low-contrast features.
RG • Volume 30 Number 2
von Falck et al 325
Figure 8. Benefits of combining thin-section
scanning with sliding-thin-slab averaging for
CT evaluation of the kidneys. (a, b) Coronal
reformatted images from a raw volume data set
reconstructed with section thicknesses of 1
mm (a) and 5 mm (b) show higher spatial resolution in a but less image noise in b. (c) Coronal
image obtained by applying the sliding-thin-slab
algorithm to the reconstructed data set used in
a shows an improvement in image noise, with
preservation of high through-plane resolution.
Figures 9, 10. (9) Coronal (left column) and axial (right column) images reconstructed from the same
thin-section CT image data set by using the standard algorithm with a section thickness of 1 mm (top
row) and the sliding-thin-slab averaging algorithm with a slab thickness of 5 mm (bottom row) show
differences in noise characteristics related to the plane and thickness of sections. Noise is less uniform
with greater low-frequency content on the coronal reformatted images than on the axial images, a characteristic that makes lesion detection more challenging. As expected, both images reconstructed with
sliding-thin-slab averaging show improvement in the SNR, with a more pronounced effect on the coronal image. (10) Graph shows the relationship between image noise (calculated as the standard deviation
of gray-scale values, from 0 to 12 au) and slab thickness (number of acquired sections included in the
slab, from one to 20 sections) for axial (solid line) and coronal (dotted line) reconstructions. Note that
image noise is uniformly higher on axial images than on coronal views.
326 March-April 2010
The combination of thin-section acquisitions
with thick reconstructions is a useful paradigm
for multidetector CT with dose-saving protocols.
References
1.Rubin GD. 3-D imaging with MDCT. Eur J Radiol
2003;45(suppl 1):S37–S41.
2.Dalrymple NC, Prasad SR, Freckleton MW,
Chintapalli KN. Introduction to the language of
three-dimensional imaging with multidetector CT.
RadioGraphics 2005;25(5):1409–1428.
3.Rydberg J, Sandrasegaran K, Tarver RD, Frank MS,
Conces DJ, Choplin RH. Routine isotropic computed tomography scanning of chest: value of coronal
and sagittal reformations. Invest Radiol 2007;42(1):
23–28.
4.Sandrasegaran K, Rydberg J, Lall CG, Hameed T,
Hawes DR, Kopecky KK. Routine isotropic computed tomography scanning of the abdomen and
pelvis. Australas Radiol 2006;50(2):93–101.
5.Das M, Mühlenbruch G, Katoh M, et al. Automated
volumetry of solid pulmonary nodules in a phantom:
accuracy across different CT scanner technologies.
Invest Radiol 2007;42(5):297–302.
6.Abdelmoumene A, Chevallier P, Chalaron M, et al.
Detection of liver metastases under 2 cm: comparison of different acquisition protocols in four row
multidetector-CT (MDCT). Eur Radiol 2005;15
(9):1881–1887.
7.Kawata S, Murakami T, Kim T, et al. Multidetector
CT: diagnostic impact of slice thickness on detection
of hypervascular hepatocellular carcinoma. AJR Am
J Roentgenol 2002;179(1):61–66.
8.Prokop M. Multislice CT: technical principles
and future trends. Eur Radiol 2003;13(suppl 5):
M3–M13.
9.Buzug TM. Image quality and artefacts. In: From
photon statistics to modern cone-beam CT. Heidelberg, Germany: Springer, 2008; 403–484.
10.Shin HO, Falck CV, Galanski M. Low-contrast detectability in volume rendering: a phantom study on
multidetector-row spiral CT data. Eur Radiol 2004;
14(2):341–349.
11.von Falck C, Hartung A, Berndzen F, King B,
Galanski M, Shin HO. Optimization of low-contrast
detectability in thin-collimated modern multidetector CT using an interactive sliding-thin-slab averaging algorithm. Invest Radiol 2008;43(4):229–235.
12.Brooks RA, Di Chiro G. Statistical limitations in
x-ray reconstructive tomography. Med Phys 1976;3
(4):237–240.
13.Boedeker KL, Cooper VN, McNitt-Gray MF. Application of the noise power spectrum in modern diagnostic MDCT: part I. Measurement of noise power
spectra and noise equivalent quanta. Phys Med Biol
2007;52(14):4027–4046.
14.Boedeker KL, McNitt-Gray MF. Application of the
noise power spectrum in modern diagnostic MDCT:
part II. Noise power spectra and signal to noise. Phys
Med Biol 2007;52(14):4047–4061.
15.ASTM International. Standard test method for
measurement of computed tomography (CT) sys-
radiographics.rsna.org
tem performance. http://www.astm.org/Standards
/E1695.htm. Accessed March 2, 2008.
16.Chao EH, Toth TL, Bromberg NB, et al. A statistical method of defining low contrast detectability
[abstr]. Radiology 2000;217(P):162.
17.Eckert MP, Bradley AP. Perceptual quality metrics
applied to still image compression. Signal Process
1998;70:177–200.
18.Wang Z, Bovik AC, Sheikh HR, Simoncelli EP. Image quality assessment: from error visibility to structural similarity. IEEE Trans Image Process 2004;13
(4):600–612.
19.Dalrymple NC, Prasad SR, El-Merhi FM, Chintapalli KN. Price of isotropy in multidetector CT.
RadioGraphics 2007;27(1):49–62.
20.Nagel HD. Radiation exposure in computed tomography: fundamentals, influencing parameters, dose
assessment optimization. 4th ed. Hamburg, Germany: CTB Publications, 2002.
21.Husstedt H, Prokop M, Becker H. Window width as
a dosage-relevant factor in high-contrast structures
in CT [in German]. Rofo 1998;168(2):139–143.
22.Goldman LW. Principles of CT: radiation dose and
image quality. J Nucl Med Technol 2007;35(4):213–
225.
23.Mayo-Smith WW, Gupta H, Ridlen MS, Brody JM,
Clements NC, Cronan JJ. Detecting hepatic lesions:
the added utility of CT liver window settings. Radiology 1999;210(3):601–604.
24.Pomerantz SM, White CS, Krebs TL, et al. Liver
and bone window settings for soft-copy interpretation of chest and abdominal CT. AJR Am J Roentgenol 2000;174(2):311–314.
25.Hilts M, Duzenli C. Image filtering for improved
dose resolution in CT polymer gel dosimetry. Med
Phys 2004;31(1):39–49.
26.van der Molen AJ, Geleijns J. Overranging in multisection CT: quantification and relative contribution
to dose—comparison of four 16-section CT scanners. Radiology 2007;242(1):208–216.
27.Theocharopoulos N, Damilakis J, Perisinakis K,
Gourtsoyiannis N. Energy imparted-based estimates
of the effect of z overscanning on adult and pediatric
patient effective doses from multi-slice computed
tomography. Med Phys 2007;34(4):1139–1152.
28.Turlington JZ, Higgins WE. New techniques for efficient sliding thin-slab volume visualization. IEEE
Trans Med Imaging 2001;20(8):823–835.
29.Napel S, Rubin GD, Jeffrey RB Jr. STS-MIP: a new
reconstruction technique for CT of the chest. J
Comput Assist Tomogr 1993;17(5):832–838.
30.Remy-Jardin M, Remy J, Gosselin B, Copin MC,
Wurtz A, Duhamel A. Sliding thin slab, minimum
intensity projection technique in the diagnosis of
emphysema: histopathologic-CT correlation. Radiology 1996;200(3):665–671.
31.Lee KH, Kim YH, Hahn S, et al. Computed tomography diagnosis of acute appendicitis: advantages
of reviewing thin-section datasets using sliding slab
average intensity projection technique. Invest Radiol
2006;41(7):579–585.
32.Lee KH, Hong H, Hahn S, Kim B, Kim KJ, Kim
YH. Summation or axial slab average intensity projection of abdominal thin-section CT datasets: can
they substitute for the primary reconstruction from
raw projection data? J Digit Imaging 2008;21(4):
422–432.
RG
Volume 30 • March April Issue 2009
von Falck et al
Sliding-Thin-Slab Averaging for Improved Depiction of LowContrast Lesions with Radiation Dose Savings at Thin-Section
CT
Christian von Falck, MD • Michael Galanski, MD, PhD • Hoen-oh Shin, MD
RadioGraphics 2010; 30:317–326 • Published online 10.1148/rg.302096007 • Content Codes:
Page 318
In the initial stage of data postprocessing, a secondary “raw” data set is reconstructed from a CT
volume data set that consists of overlapping sections with a thickness of 1 mm or less, acquired at
standard radiation dose levels comparable to those for scanning with thicker sections (eg, 5 mm) (8).
A relatively high degree of noise is acceptable on the images from this initial reconstruction.
Page 322
the use of a sliding-thin-slab averaging algorithm at image reconstruction allows the advantage of
thin-section scanning, namely the reduction of through-plane partial-volume averaging effects, to be
retained, while the detectability of low-contrast objects is improved by the retrospective generation of
thicker sections.
Page 324
noise is even more effectively suppressed on sagittal and coronal images than on axial images
reconstructed with the sliding-thin-slab averaging algorithm
Page 324
it was demonstrated that the optimal slab thickness depends on the size of the lesion to be detected
and that the slab thickness should be interactively varied during the reading process at the workstation
(11). Optimal lesion detectability is obtained with a slab thickness that is one-half the lesion diameter.
Page 326
The combination of thin-section acquisitions with thick reconstructions is a useful paradigm for
multidetector CT with dose-saving protocols.