DOI: 10.1148/rg.271065037
RadioGraphics 2007;27:49-62
© RSNA, 2007
Price of Isotropy in Multidetector CT1
Neal C. Dalrymple, MD,
Srinivasa R. Prasad, MD,
Fadi M. El-Merhi, MD and
Kedar N. Chintapalli, MD
1 From the Department of Radiology, University of Texas Health Science Center at San Antonio, Mail Code 7800, 7703 Floyd Curl Dr, San Antonio, TX 78229-3900. Presented as an education exhibit at the 2005 RSNA Annual Meeting. Received March 21, 2006; revision requested June 5 and received July 10; accepted July 11. All authors have no financial relationships to disclose.
Address correspondence to N.C.D. (e-mail: dalrymplen{at}uthscsa.edu).
 |
Abstract
|
|---|
Recent advances in multidetector computed tomography (CT) have made isotropic data acquisition feasible for nearly every application. The benefits of routine use of isotropic data for image display and interpretation have been described in the literature and at educational conferences. However, there is usually a trade-off in the form of an increased radiation dose to the patient. The parameters that affect the radiation dose vary considerably in accordance with the CT scanner design, and those variations determine the cost in dose increase relative to the voxel size. The detector configuration and beam collimation (narrow or wide) used for a particular acquisition also affect the voxel size and the relationship between spatial resolution and the radiation dose. By closely comparing the quality of multidetector CT images obtained with different detector configurations on scanners with four, 16, 40, and 64 channels and the estimated radiation exposure incurred with each option, radiologists may achieve an understanding of the relationship between radiation dose and voxel size. This understanding, in turn, may help balance the need for diagnostic image quality against the concern for patient safety.
© RSNA, 2007
|
Introduction
|
|---|
The introduction of computed tomography (CT) more than 3 decades ago revolutionized medical imaging. The display of axial sections dramatically improved differentiation among tissue types within the body. As CT technology has evolved, so have postprocessing applications that allow the transformation of axial CT data into nonaxial images. Such applications include multiplanar reformation and maximum intensity (slab) projection for the reconstruction of two-dimensional images and volume rendering for the creation of three-dimensional images. The accuracy of images obtained with these postprocessing methods depends on the spatial resolution of image data acquired along the long axis of the patient (ie, longitudinal, through-plane, or z-axis spatial resolution). The long axis can be imagined as a line that courses in a craniocaudal direction through a patient positioned horizontally in a scanner. Axial images, therefore, consist of data reconstructed in axial sections along the x-and y-axes.
Through several generations of CT scanners, long-axis resolution was consistently inferior to short-axis, or axial, spatial resolution. Spatial resolution in the axial plane is defined by pixel size. Within a matrix of 512 x 512 and a field of view of 25 cm, the pixels that constitute each axial image are squares with a length of approximately 0.49 mm on each side. If the field of view is increased to 40 cm, the length of each pixel is approximately 0.78 mm on each side. Since spatial resolution in the longitudinal plane is dependent on section thickness, a section thickness in the range of 0.5–0.8 mm is required to achieve similar spatial resolution in all three dimensions (1). If the thickness of the axial section is taken into account, the square pixels are converted to three-dimensional voxels. When data are reconstructed to achieve similar dimensions in all three planes, the data are considered to be isotropic (2). Isotropic data consist of cube-shaped voxels of equal length on each side (Fig 1). On the other hand, if the reconstructed section thickness is greater than the pixel size, the data are considered anisotropic, and spatial resolution in the longitudinal plane is inferior to that in the transverse plane.
View larger version (36K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a. Geometry of isotropic and anisotropic acquisitions. Anisotropic data consist of voxels that have a section thickness greater than the x- and y-axis dimensions of the facing pixels. Section thickness along the z-axis is four times the size of each pixel in a but only twice the size of each pixel in b. Although both data sets are anisotropic, there is a significant difference in image quality for three-dimensional applications, with improved longitudinal spatial resolution in b compared with that in a. When the section thickness is equal to the pixel size, as in c, the data are isotropic.
|
|
View larger version (38K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b. Geometry of isotropic and anisotropic acquisitions. Anisotropic data consist of voxels that have a section thickness greater than the x- and y-axis dimensions of the facing pixels. Section thickness along the z-axis is four times the size of each pixel in a but only twice the size of each pixel in b. Although both data sets are anisotropic, there is a significant difference in image quality for three-dimensional applications, with improved longitudinal spatial resolution in b compared with that in a. When the section thickness is equal to the pixel size, as in c, the data are isotropic.
|
|
View larger version (41K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1c. Geometry of isotropic and anisotropic acquisitions. Anisotropic data consist of voxels that have a section thickness greater than the x- and y-axis dimensions of the facing pixels. Section thickness along the z-axis is four times the size of each pixel in a but only twice the size of each pixel in b. Although both data sets are anisotropic, there is a significant difference in image quality for three-dimensional applications, with improved longitudinal spatial resolution in b compared with that in a. When the section thickness is equal to the pixel size, as in c, the data are isotropic.
|
|
Recent advances in multidetector CT technology have made the acquisition of isotropic data feasible in nearly every CT examination with use of a narrow configuration of the detector array so that only the smallest detector elements are exposed. The benefits of routine acquisitions of isotropic data have been described in the literature and at educational conferences (3,4). However, there is usually a trade-off in the form of an increased radiation dose to the patient. In addition, the use of a narrow detector collimation also prolongs the scanning time, a factor that affects the coverage that may be achieved within a breath hold or with an intravenous contrast medium bolus. The parameters that affect the radiation dose and exposure time vary considerably according to scanner design, and these variations determine the proportions of the trade-off in increased radiation dose and scanning time relative to the voxel size. An understanding of the relationships between voxel size, image quality, and radiation dose allows the radiologist to balance the need for high-quality diagnostic images against the concern for maintaining patient safety by minimizing the radiation dose in accordance with the ALARA (as low as reasonably achievable) principle (5–7).
To achieve a better understanding of the effects of variation among these multidetector CT parameters, we examined the relationship between voxel geometry and estimated radiation dose with the use of four-, 16-, 40-, and 64-channel CT scanners. The cases chosen for inclusion in this article allow comparisons of image quality on the basis of voxel size and the corresponding relative radiation dose. One patient underwent several CT examinations for clinical reasons, and the use of different detector configurations permitted the reconstruction of sections of various thicknesses. In other cases, the images provided for comparison were created by using multiple data reconstructions from a single CT examination. For example, the acquisition of data with a 16-channel scanner and a 16 x 0.625 mm detector configuration (ie, with all 16 channels active) allows reconstruction both with isotropic voxels (with a section thickness of 0.625 mm) and with anisotropic voxels (with a section thickness of 1.25 mm). Anisotropic data reconstruction provides information similar to that obtainable with image acquisition performed with a wider collimation (eg, the 16 x 1.25 mm detector configuration). This method of reconstruction was used because it is difficult to control multiple variables when scanning the same patient on multiple occasions. Therefore, the detector configuration listed for each case included in this article is the widest section collimation setting that could provide the data represented (except in one case, in which several image data sets from separate acquisitions were available). For each data set, a reconstruction increment was selected to provide a 50% overlap.
Comparable radiation dose estimates were obtained for separate examinations by prospectively planning the use of different detector configurations and recording the available (weighted or volume) CT dose index values. For CT examinations performed with the four-channel scanner, which was not equipped with automated tube current modulation, a constant tube current value was used. With the 16-, 40-, and 64-channel scanners, automated current modulation was available and was used for virtually all clinical examinations, so a fixed noise index or tube current per section was selected, depending on the scanner.
In all cases, the selected reconstruction increment was 50% of section thickness, to provide overlapping sections. Although the minimum section thickness achievable at data reconstruction is determined by the effective detector width at the time of acquisition, the increment for data reconstruction is arbitrary and may be thinner than the section. Thus, overlapping data reconstruction may be used to optimize the z-axis resolution for a given acquisition (8–11). For example, the quality of a maximum intensity projection image from CT data reconstructed with a 1.25-mm section thickness is optimal when the sections are reconstructed with an increment of 0.625 mm. All of the images in the present article were reconstructed in this way, with section overlap.
It should be noted that the discussion of "price" in this article does not refer to the cost of the scanner or of its use. An "expensive" scanner platform, therefore, is a CT system that requires a considerable increase in the radiation dose to the patient (usually accompanied by prolonged scanning time and limited z-axis coverage) to achieve isotropic or near-isotropic voxel dimensions. Thus, the newest scanners, which have the highest monetary price tag, are desirable precisely because they enable the achievement of isotropy with minimal trade-offs in radiation dose and acquisition time. The "cost" of time also must be kept in perspective. A decrease in scanning time from 40 to 10 seconds is unlikely to affect the overall time the patient spends in the CT suite but has tremendous effects on the amount of volume coverage achievable within a particular breath hold or with a particular contrast medium bolus, as well as on the effectiveness of cardiac gating.
The investigative method we used does not permit an accurate comparison of the amounts of overall radiation exposure incurred with the various scanner platforms. Many confounding variables exist, including different scanning protocols (with different tube current settings) for each platform, the lack of automated current modulation on our four-channel scanner, and the variable accuracy of estimated CT dose index values furnished by different vendors. The selected cases included in this article only indicate the relative influence of various possible detector configurations within each multidetector CT platform.
|
Basic Principles of Multidetector CT
|
|---|
Multidetector CT has revolutionized the versatility of CT data with the partitioning of information received from the incident beam into multiple small segments (12). In many cases, the reconstructed sections used for interpretation consist of combined data from several detector elements. If two or more detector elements are combined, or binned, by the data acquisition system at the level of the detector array, the combined size of the two elements determines the effective section thickness and, thus, the thinnest section that can be reconstructed.
The term detector configuration describes the number of data collection channels and the effective section thickness determined by the data acquisition system settings. Figure 2 shows the beam geometry on a 16-channel scanner with two different detector configurations—16 x 0.625 mm and 16 x 1.25 mm (2). Depending on the setting used, both the size and the geometric characteristics of the incident x-ray beam change. The changes in beam geometry and their implications for dose efficiency and scanning time affect the radiation dose to the patient.
View larger version (29K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a. General principles of detector configuration at multidetector CT with a 16-channel scanner. (a) Diagram of beam geometry with the 16 x 0.625 mm detector configuration shows narrow collimation of the x-ray beam, with exposure of only the central detector elements (DE). The data acquisition system (DAS), a type of switchable circuit system, is set to sample each of the central elements individually. This setting permits reconstruction of 0.625-mm-thick sections, the thinnest possible with this scanner platform. Note that a portion of the beam (rose bands) extends beyond the active detector elements. This area of overextension, called the penumbra, is necessary to ensure exposure of the most peripheral of the active detector elements, but beam overextension results in some radiation exposure that does not directly contribute to the image. (b) Diagram of beam geometry with the 16 x 1.25 mm detector configuration shows wider collimation of the x-ray beam to expose all the detector elements. The data acquisition system samples combined data from the small central elements while collecting data separately from each of the larger peripheral elements. In this setting, a larger volume of tissue is exposed per gantry rotation, but axial sections cannot be reconstructed to a thickness of less than 1.25 mm. Because the penumbra is a smaller percentage of the overall beam, the efficiency of the acquisition is increased and the patient is exposed to less additional radiation.
|
|
View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b. General principles of detector configuration at multidetector CT with a 16-channel scanner. (a) Diagram of beam geometry with the 16 x 0.625 mm detector configuration shows narrow collimation of the x-ray beam, with exposure of only the central detector elements (DE). The data acquisition system (DAS), a type of switchable circuit system, is set to sample each of the central elements individually. This setting permits reconstruction of 0.625-mm-thick sections, the thinnest possible with this scanner platform. Note that a portion of the beam (rose bands) extends beyond the active detector elements. This area of overextension, called the penumbra, is necessary to ensure exposure of the most peripheral of the active detector elements, but beam overextension results in some radiation exposure that does not directly contribute to the image. (b) Diagram of beam geometry with the 16 x 1.25 mm detector configuration shows wider collimation of the x-ray beam to expose all the detector elements. The data acquisition system samples combined data from the small central elements while collecting data separately from each of the larger peripheral elements. In this setting, a larger volume of tissue is exposed per gantry rotation, but axial sections cannot be reconstructed to a thickness of less than 1.25 mm. Because the penumbra is a smaller percentage of the overall beam, the efficiency of the acquisition is increased and the patient is exposed to less additional radiation.
|
|
|
Near Isotropy with Four-Channel CT Scanners
|
|---|
Near-isotropic acquisitions on any four-channel scanner exact a hefty penalty in terms of increased radiation dose and scanning time. There is considerable variation in detector design among four-channel CT scanners. Matrix arrays consist of detector elements of identical size. Adaptive arrays consist of elements of several different sizes, with the smallest elements positioned at the center of the array, and with the sizes of the other elements progressively increasing toward the periphery of the array. Hybrid arrays consist of two sizes of detector elements, with small elements in the central region and larger ones toward the periphery. Despite the differences in scanner design, the principles of binning detector elements and changing beam geometry are generally applicable to all multidetector systems. For the sake of simplicity, the discussion in the following paragraphs is focused on the configuration of a four-channel CT scanner with a matrix array.
A typical four-channel multidetector CT scanner offers four potential detector configurations. One of these four settings must be used for any helical acquisition, regardless of the specified section thickness for data reconstruction. The detector configurations available on a matrix-array four-channel scanner are shown in Figure 3. The matrix array consists of 16 detector elements, each with a diameter of 1.25 mm. Exposure of four central elements results in four data components with a section thickness of 1.25 mm each. Exposure of eight central elements results in four data components with a section thickness of 2.5 mm each. The detectors in the matrix array also can be configured for the acquisition of four data components with a section thickness of 3.75 mm or 5.0 mm each. Representative images obtained with each of these settings are shown in Figures 4 and 5.
View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a. Detector configurations and minimal voxel dimensions at multidetector CT with a matrix-array four-channel scanner. (a) Left: Diagram of beam geometry with a detector configuration of 4 x 1.25 mm. Four central elements are exposed, and four data components with a section thickness of 1.25 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed near-isotropic voxels, which on axial images have a z-axis depth of 1.25 mm, approximately twice the x- and y-axis dimension (0.7 mm). (b) Left: Diagram of beam geometry with a detector configuration of 4 x 2.5 mm. Eight central elements are exposed in pairs, and four data components with a section thickness of 2.5 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are four times as long in the z-axis as in the x- and y-axes. (c) Left: Diagram of beam geometry with a detector configuration of 4 x 3.75 mm. Twelve central elements are exposed in triplets, and four data components with a section thickness of 3.75 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are approximately six times as long in the z-axis as in the x- and y-axes. (d) Left: Diagram of beam geometry with a detector configuration of 4 x 5.0 mm. All 16 detector elements are exposed, and four data components with a section thickness of 5.0 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are eight times as long in the z-axis as in the x- and y-axes.
|
|
View larger version (27K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b. Detector configurations and minimal voxel dimensions at multidetector CT with a matrix-array four-channel scanner. (a) Left: Diagram of beam geometry with a detector configuration of 4 x 1.25 mm. Four central elements are exposed, and four data components with a section thickness of 1.25 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed near-isotropic voxels, which on axial images have a z-axis depth of 1.25 mm, approximately twice the x- and y-axis dimension (0.7 mm). (b) Left: Diagram of beam geometry with a detector configuration of 4 x 2.5 mm. Eight central elements are exposed in pairs, and four data components with a section thickness of 2.5 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are four times as long in the z-axis as in the x- and y-axes. (c) Left: Diagram of beam geometry with a detector configuration of 4 x 3.75 mm. Twelve central elements are exposed in triplets, and four data components with a section thickness of 3.75 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are approximately six times as long in the z-axis as in the x- and y-axes. (d) Left: Diagram of beam geometry with a detector configuration of 4 x 5.0 mm. All 16 detector elements are exposed, and four data components with a section thickness of 5.0 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are eight times as long in the z-axis as in the x- and y-axes.
|
|
View larger version (30K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3c. Detector configurations and minimal voxel dimensions at multidetector CT with a matrix-array four-channel scanner. (a) Left: Diagram of beam geometry with a detector configuration of 4 x 1.25 mm. Four central elements are exposed, and four data components with a section thickness of 1.25 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed near-isotropic voxels, which on axial images have a z-axis depth of 1.25 mm, approximately twice the x- and y-axis dimension (0.7 mm). (b) Left: Diagram of beam geometry with a detector configuration of 4 x 2.5 mm. Eight central elements are exposed in pairs, and four data components with a section thickness of 2.5 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are four times as long in the z-axis as in the x- and y-axes. (c) Left: Diagram of beam geometry with a detector configuration of 4 x 3.75 mm. Twelve central elements are exposed in triplets, and four data components with a section thickness of 3.75 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are approximately six times as long in the z-axis as in the x- and y-axes. (d) Left: Diagram of beam geometry with a detector configuration of 4 x 5.0 mm. All 16 detector elements are exposed, and four data components with a section thickness of 5.0 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are eight times as long in the z-axis as in the x- and y-axes.
|
|
View larger version (31K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3d. Detector configurations and minimal voxel dimensions at multidetector CT with a matrix-array four-channel scanner. (a) Left: Diagram of beam geometry with a detector configuration of 4 x 1.25 mm. Four central elements are exposed, and four data components with a section thickness of 1.25 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed near-isotropic voxels, which on axial images have a z-axis depth of 1.25 mm, approximately twice the x- and y-axis dimension (0.7 mm). (b) Left: Diagram of beam geometry with a detector configuration of 4 x 2.5 mm. Eight central elements are exposed in pairs, and four data components with a section thickness of 2.5 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are four times as long in the z-axis as in the x- and y-axes. (c) Left: Diagram of beam geometry with a detector configuration of 4 x 3.75 mm. Twelve central elements are exposed in triplets, and four data components with a section thickness of 3.75 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are approximately six times as long in the z-axis as in the x- and y-axes. (d) Left: Diagram of beam geometry with a detector configuration of 4 x 5.0 mm. All 16 detector elements are exposed, and four data components with a section thickness of 5.0 mm each are acquired per gantry rotation. Right: Diagram shows the reconstructed voxels, which are eight times as long in the z-axis as in the x- and y-axes.
|
|
View larger version (95K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4a. Coronal multidetector CT images of the right hip, obtained with various detector configurations on a four-channel scanner. In each case, a fixed tube current of 240 mA was used, and gantry rotation time and table speed were constant. Data were reconstructed with an overlap and with each section thickness available with the particular detector configuration used. The reconstructed section thickness and increment were 5.0 mm and 2.5 mm (a), 3.75 mm and 1.8 mm (b), 2.5 mm and 1.25 mm (c), and 1.25 mm and 0.625 mm (d). Note that reconstruction with the thinnest section available is required for adequate evaluation of the articular cortex and joint space of the hip. Config. = detector configuration, CTDIw = weighted CT dose index, Time = scanning time.
|
|
View larger version (98K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4b. Coronal multidetector CT images of the right hip, obtained with various detector configurations on a four-channel scanner. In each case, a fixed tube current of 240 mA was used, and gantry rotation time and table speed were constant. Data were reconstructed with an overlap and with each section thickness available with the particular detector configuration used. The reconstructed section thickness and increment were 5.0 mm and 2.5 mm (a), 3.75 mm and 1.8 mm (b), 2.5 mm and 1.25 mm (c), and 1.25 mm and 0.625 mm (d). Note that reconstruction with the thinnest section available is required for adequate evaluation of the articular cortex and joint space of the hip. Config. = detector configuration, CTDIw = weighted CT dose index, Time = scanning time.
|
|
View larger version (99K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4c. Coronal multidetector CT images of the right hip, obtained with various detector configurations on a four-channel scanner. In each case, a fixed tube current of 240 mA was used, and gantry rotation time and table speed were constant. Data were reconstructed with an overlap and with each section thickness available with the particular detector configuration used. The reconstructed section thickness and increment were 5.0 mm and 2.5 mm (a), 3.75 mm and 1.8 mm (b), 2.5 mm and 1.25 mm (c), and 1.25 mm and 0.625 mm (d). Note that reconstruction with the thinnest section available is required for adequate evaluation of the articular cortex and joint space of the hip. Config. = detector configuration, CTDIw = weighted CT dose index, Time = scanning time.
|
|
View larger version (101K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4d. Coronal multidetector CT images of the right hip, obtained with various detector configurations on a four-channel scanner. In each case, a fixed tube current of 240 mA was used, and gantry rotation time and table speed were constant. Data were reconstructed with an overlap and with each section thickness available with the particular detector configuration used. The reconstructed section thickness and increment were 5.0 mm and 2.5 mm (a), 3.75 mm and 1.8 mm (b), 2.5 mm and 1.25 mm (c), and 1.25 mm and 0.625 mm (d). Note that reconstruction with the thinnest section available is required for adequate evaluation of the articular cortex and joint space of the hip. Config. = detector configuration, CTDIw = weighted CT dose index, Time = scanning time.
|
|
View larger version (147K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5a. Volume-rendered images obtained with different section thicknesses at four-channel CT. The first row of data on the images is the detector configuration, the middle row is the weighted CT dose index, and the bottom row is the scanning time. (a) Image obtained with a reconstructed section thickness of 5.0 mm and an increment of 2.5 mm shows the right hepatic artery (arrow) where it arises from the celiac artery. The origin of the left hepatic artery (arrowhead) at the left gastric artery also is visible but is not well depicted. (b) Image in a similar orientation, obtained with a reconstructed section thickness of 2.5 mm and an increment of 1.25 mm, shows both hepatic arteries more clearly and with a smoother appearance than in a.
|
|
View larger version (160K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5b. Volume-rendered images obtained with different section thicknesses at four-channel CT. The first row of data on the images is the detector configuration, the middle row is the weighted CT dose index, and the bottom row is the scanning time. (a) Image obtained with a reconstructed section thickness of 5.0 mm and an increment of 2.5 mm shows the right hepatic artery (arrow) where it arises from the celiac artery. The origin of the left hepatic artery (arrowhead) at the left gastric artery also is visible but is not well depicted. (b) Image in a similar orientation, obtained with a reconstructed section thickness of 2.5 mm and an increment of 1.25 mm, shows both hepatic arteries more clearly and with a smoother appearance than in a.
|
|
As mentioned earlier, significant trade-offs must be made to achieve thin-section acquisitions with a four-channel multidetector CT scanner. With the narrowest section collimation (4 x 1.25 mm detector configuration), the radiation dose the patient may be twice that incurred with the widest collimation (4 x 5.0 mm). Our four-channel scanner is not equipped with automated tube current modulation, and the current is kept constant, so the radiation dose increase is not incurred in compensation for the avoidance of image noise. Rather, it results from overextension of the x-ray beam, which is inherent in multidetector CT technology. 
To ensure adequate data collection by the most peripheral detector elements, the x-ray beam must extend beyond the active elements, and this overextension of the beam produces a penumbra effect at the margins of the detector array (7). Beam overextension also results in wasted radiation that does not contribute directly to the image. Since the penumbra accounts for a larger percentage of the beam when a narrow detector configuration is used, dose efficiency is decreased with the use of a narrow collimation. This principle applies to CT performed with any multidetector CT scanner, but the effect on radiation dose is greater with four-channel scanners (because of the small beam size) than with scanners that have a higher number of channels. Although this effect is not necessarily prohibitive, it does imply that there should be a demonstrable need for thin-section acquisition. Short segments of small vessels may be more visible at CT angiography with a narrow collimation. In addition, the depiction of fine osseous detail may be important for evaluation of the spine in a trauma setting.
Like the radiation dose, the long-axis coverage per gantry rotation may vary, with as much as a fourfold difference between narrow and wide detector configurations. The detector configuration has significant implications for the volume of tissue that may be imaged in a particular breath hold or with a specific bolus of an intravenous contrast medium. When narrow collimation is used at CT angiography, a large volume of contrast medium may be needed, and bolus timing is critical to avoid a loss of contrast during the examination. Prolonged exposure time also limits the ability to perform multiphase examinations, such as liver imaging with arterial and portal venous phases. Although the liver can usually be imaged effectively during the arterial phase, there may not be enough time for the necessary tube cooling before scanning during the portal venous phase. For that reason, we perform dynamic liver and renal imaging with a 4 x 2.5 mm detector configuration.
|
Isotropy with 16-Channel CT Scanners
|
|---|
The introduction of multidetector CT scanners with 16 channels revolutionized the perception of CT data. Whereas the limited long-axis coverage provided by four-channel CT scanners in the narrow collimation mode restricts thin-section acquisitions to small scanning volumes or short scanning lengths that do not require a breath hold, 16-channel scanners allow a section thickness approaching 1 mm for virtually every examination. Thus, thin-section data reconstruction need not be anticipated at scanning. On 16-channel scanners, isotropy is achievable with a lesser penalty in increased radiation dose and with somewhat shorter scanning time than with four-channel scanners.
Many different types of 16-channel CT scanners are available, and the detector design differs among vendors; however, the geometric principles of detector design differ considerably less among scanners with 16 channels than among those with four channels. In general, there are only two possible detector configurations for most 16-channel multidetector CT examinations (Fig 6). For example, collimation of the incident beam so that only the central 16 detectors are exposed results in the acquisition of 16 sections with a thickness of 0.625 mm each per gantry rotation. This may be considered the narrow collimation mode. If the collimator is adjusted so that the full detector array is exposed to the incident beam, then 16 sections with a thickness of 1.25 mm each are acquired per gantry rotation. This may be considered the wide collimation mode.
View larger version (34K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6a. Detector configurations and voxel dimensions at 16-channel multidetector CT. (a) Left: Diagram shows narrow collimation, with exposure of only the central 16 detector elements. Each element functions as a separate unit, and 16 sections with a thickness of 0.625 mm each are acquired per gantry rotation. Right: Diagram shows that the reconstructed voxels are isotropic, with about equal length in each dimension. (b) Left: Diagram shows wide collimation, with exposure not only of the central small elements but also of larger elements at the periphery. Central elements function in pairs, and peripheral elements are used individually. As a result, 16 sections with a thickness of 1.25 mm each are acquired per rotation. Right: Diagram shows that reconstructed voxels are anisotropic, about twice as long in the longitudinal plane as in the transverse plane.
|
|
View larger version (34K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6b. Detector configurations and voxel dimensions at 16-channel multidetector CT. (a) Left: Diagram shows narrow collimation, with exposure of only the central 16 detector elements. Each element functions as a separate unit, and 16 sections with a thickness of 0.625 mm each are acquired per gantry rotation. Right: Diagram shows that the reconstructed voxels are isotropic, with about equal length in each dimension. (b) Left: Diagram shows wide collimation, with exposure not only of the central small elements but also of larger elements at the periphery. Central elements function in pairs, and peripheral elements are used individually. As a result, 16 sections with a thickness of 1.25 mm each are acquired per rotation. Right: Diagram shows that reconstructed voxels are anisotropic, about twice as long in the longitudinal plane as in the transverse plane.
|
|
Since geometric efficiency improves with collimation greater than 10 mm, the geometry of the gantry design on 16-channel scanners produces a less marked difference in the beam overextension effect between wide and narrow detector configurations than does the geometry of the gantry design on four-channel scanners (13). In theory, the greater geometric efficiency of 16-channel scanners should result in a lesser difference in radiation dose to the patient with a wide versus a narrow collimation setting. However, this hypothesis is not supported by our experience in clinical practice, perhaps at least in part because of a compensatory increase in tube current directed by the automated current modulation program to minimize image noise from the smaller detector elements by increasing photon flux.
Automated current modulation programs became widely available around the time when 16-channel scanners were introduced. Thus, many institutions encountered both multidetector CT and tube current modulation for the first time. The current may be modulated within a given gantry rotation to vary the dose in the x- and y-planes (transverse modulation) as well as along the long axis of the patient (z-axis modulation). Although automated current modulation programs are intended to optimize the use of tube current so as to maximize the image quality while minimizing the radiation dose to the patient, the interaction of such programs with variable scanning parameters can be complicated (14,15). At a specified level of tube output, thin-section acquisitions incur higher levels of noise than do acquisitions of thicker sections. Thus, tube current modulation programs that self-regulate according to a user-specified level of image noise will increase the tube current when a lesser section thickness is selected, compared with the tube current when thicker sections are specified. When a 16-channel scanner is used, the selection of section thickness may have at least as great an effect on radiation dose to the patient as does the geometric efficiency of the detector configuration. Some may choose to avoid this pitfall by selecting relatively thick sections for primary axial interpretation (and adjusting the settings to optimize the noise level of those sections) and thin sections for retrospective reconstruction of multiplanar reformatted images and volume-rendered images. This method helps preserve the image quality of the axial sections but results in increased noise on the processed images.
Although isotropic voxels are achieved only with the narrow detector configuration, the anisotropic voxels available with 16-channel scanners in the wide collimation mode provide long-axis resolution that suffices for many applications. With a 16-channel platform, narrow collimation provides half the volume coverage achievable with wide collimation, and an increased radiation dose. Although isotropic acquisitions are routinely achievable with 16-channel multidetector CT, the excellent multiplanar image quality available with the wider collimation setting makes the routine use of narrow collimation questionable (Figs 7, 8). A comparison of three-dimensional and multiplanar reformatted images obtained with wide versus narrow collimation on a 16-channel scanner allows an informed consideration of whether the difference in image quality with the use of narrow collimation for a particular application justifies the increase in radiation dose.
View larger version (131K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7a. Depiction of a lung nodule with different detector configurations on a 16-channel CT scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Coronal reformatted image obtained with the wide detector configuration (reconstructed section thickness, 1.25 mm; increment, 0.625 mm) provides excellent depiction of a nodule and adjacent vessels in the upper lobe of the left lung. (b) Coronal reformatted image obtained with the narrow detector configuration (reconstructed section thickness, 0.625 mm; increment, 0.312 mm) provides slightly improved depiction of vessels and interstitial lines, but there is no alteration in the quality of depiction of the nodule.
|
|
View larger version (134K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7b. Depiction of a lung nodule with different detector configurations on a 16-channel CT scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Coronal reformatted image obtained with the wide detector configuration (reconstructed section thickness, 1.25 mm; increment, 0.625 mm) provides excellent depiction of a nodule and adjacent vessels in the upper lobe of the left lung. (b) Coronal reformatted image obtained with the narrow detector configuration (reconstructed section thickness, 0.625 mm; increment, 0.312 mm) provides slightly improved depiction of vessels and interstitial lines, but there is no alteration in the quality of depiction of the nodule.
|
|
View larger version (62K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8a. Three-dimensional volume-rendered images from renal CT angiography with a 16-channel scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Image reconstructed from anisotropic data provides satisfactory depiction of the aorta and central vessels. (b) Image reconstructed from isotropic data provides slightly improved definition of smaller vessels. The automated "seed and grow" software program used to create these images provided better depiction of peripheral branches of the renal vessels with the use of isotropic data.
|
|
View larger version (101K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8b. Three-dimensional volume-rendered images from renal CT angiography with a 16-channel scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Image reconstructed from anisotropic data provides satisfactory depiction of the aorta and central vessels. (b) Image reconstructed from isotropic data provides slightly improved definition of smaller vessels. The automated "seed and grow" software program used to create these images provided better depiction of peripheral branches of the renal vessels with the use of isotropic data.
|
|
|
Isotropy with 40-Channel CT Scanners
|
|---|
For thin-section data acquisition, 40-channel multidetector CT scanners offer the same benefits available with 16-channel scanners, but at a fraction of the penalty; isotropy is achievable with only modest increases in radiation dose and scanning time. The 40-channel scanners have two main detector configurations that are used for most helical acquisitions, 40 x 0.625 mm and 32 x 1.25 mm. With the latter, as with the narrow collimation setting on 16-channel scanners, data are acquired separately by each of the central 40 detector elements. With the wider collimation setting, the data are collected by pairs of small central detector elements as well as by larger peripheral elements (Fig 9). There is a difference in volume coverage per gantry rotation (25 mm in the narrow and 40 mm in the wide collimation mode), but volume coverage with the narrow collimation setting is comparable to or greater than that with a 16-channel scanner. Thus, the time penalty and its effects on breath-hold and contrast medium–enhanced imaging are relatively insignificant. The time cost is further reduced in this generation of scanners by the incorporation of high-output tubes with sufficient current to allow rapid rates of gantry rotation with acceptable levels of noise. Indeed, CT angiography can be performed by using a very small amount of contrast material, if the bolus timing is precise. For solid organ imaging, scanning time is no longer a significant factor in determining the necessary quantity of the contrast medium. Rather, the amount of iodine necessary for adequate parenchymal opacification determines the minimum amount of contrast medium required. Figures 10 and 11 allow a comparison of image quality and estimated radiation dose with anistropic and isotropic acquisitions on a 40-channel scanner platform.
View larger version (29K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9a. Detector configurations and voxel dimensions at 40-channel multidetector CT. (a) Left: Diagram shows narrow beam collimation, with exposure of only the central 40 detector elements. In this setting, each element functions as a separate unit, and 40 sections with a thickness of 0.625 mm each are acquired per gantry rotation, yielding long-axis coverage of 25 mm. Right: Diagram shows that reconstructed voxels in this mode are isotropic, with approximately equal length in each dimension. (b) Left: Diagram shows wide beam collimation, with exposure not only of the central small elements but also of the additional larger elements at the periphery of the array. The central elements function in pairs, and the larger peripheral elements are each used individually. As a result, 32 sections with a thickness of 1.25 mm each are acquired per gantry rotation, yielding long-axis coverage of 40 mm. Right: Diagram shows that reconstructed voxels in this mode are anisotropic, approximately twice as large in the longitudinal plane as in the transverse plane.
|
|
View larger version (28K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9b. Detector configurations and voxel dimensions at 40-channel multidetector CT. (a) Left: Diagram shows narrow beam collimation, with exposure of only the central 40 detector elements. In this setting, each element functions as a separate unit, and 40 sections with a thickness of 0.625 mm each are acquired per gantry rotation, yielding long-axis coverage of 25 mm. Right: Diagram shows that reconstructed voxels in this mode are isotropic, with approximately equal length in each dimension. (b) Left: Diagram shows wide beam collimation, with exposure not only of the central small elements but also of the additional larger elements at the periphery of the array. The central elements function in pairs, and the larger peripheral elements are each used individually. As a result, 32 sections with a thickness of 1.25 mm each are acquired per gantry rotation, yielding long-axis coverage of 40 mm. Right: Diagram shows that reconstructed voxels in this mode are anisotropic, approximately twice as large in the longitudinal plane as in the transverse plane.
|
|
View larger version (176K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 10a. Coronal reformatted images from renal CT angiography performed with a 40-channel scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Image reconstructed from anisotropic data provides excellent definition of the origins of both renal arteries and depicts both calcified and noncalcified components of plaque. (b) Image reconstructed from isotropic data provides minimally improved definition of the margins of noncalcified plaque in the left renal artery, with an increase of 20% in the estimated radiation dose to the patient. Although the scanning time required for isotropic data acquisition is 60% greater than that for anisotropic data acquisition, the difference (3 sec) is unlikely to have a significant effect on breath-hold or contrast-enhanced multiphase imaging.
|
|
View larger version (173K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 10b. Coronal reformatted images from renal CT angiography performed with a 40-channel scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Image reconstructed from anisotropic data provides excellent definition of the origins of both renal arteries and depicts both calcified and noncalcified components of plaque. (b) Image reconstructed from isotropic data provides minimally improved definition of the margins of noncalcified plaque in the left renal artery, with an increase of 20% in the estimated radiation dose to the patient. Although the scanning time required for isotropic data acquisition is 60% greater than that for anisotropic data acquisition, the difference (3 sec) is unlikely to have a significant effect on breath-hold or contrast-enhanced multiphase imaging.
|
|
View larger version (97K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 11a. Magnified views of the abdominal aorta at the level of the hepatic artery, from CT angiography performed with a 40-channel scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Volume-rendered image reconstructed from anisotropic data provides adequate depiction of a stenosis of the replaced right hepatic artery at its origin from the superior mesenteric artery (arrow). The use of automated segmentation software consistently resulted in the removal of a calcified plaque (arrowhead) at the origin of the superior mesenteric artery. (b) Volume-rendered image reconstructed from isotropic data clearly shows the plaque (arrowhead).
|
|
View larger version (122K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 11b. Magnified views of the abdominal aorta at the level of the hepatic artery, from CT angiography performed with a 40-channel scanner. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Volume-rendered image reconstructed from anisotropic data provides adequate depiction of a stenosis of the replaced right hepatic artery at its origin from the superior mesenteric artery (arrow). The use of automated segmentation software consistently resulted in the removal of a calcified plaque (arrowhead) at the origin of the superior mesenteric artery. (b) Volume-rendered image reconstructed from isotropic data clearly shows the plaque (arrowhead).
|
|
|
Isotropy with 64-Channel Scanners
|
|---|
The 64-channel scanner, which is designed to maximize z-axis coverage, is the multidetector CT platform that most recently became available commercially. The large volume of coverage per gantry rotation (32–40 mm, depending on the scanner model), coupled with an increased rate of gantry rotation, has improved the temporal resolution of CT image data significantly. With 64-channel scanners, isotropic acquisitions are routinely achievable with virtually no penalty in increased radiation dose and scanning time. The improved temporal resolution is most beneficial for cardiac imaging and may help minimize both the amount of contrast medium needed at CT angiography and the severity of motion-related artifacts.
Most of the 64-channel scanners available are based on the same general principles. The 64 detector elements may be used individually or paired by the data acquisition system to achieve the effect of 32 larger elements. Note that the size of the incident x-ray beam remains constant in both collimation modes on 64-channel scanners, unlike the size of the beam with the other scanner platforms described in this article (Fig 12). Thus, the time required for acquisition does not differ between detector configurations. In addition, since the beam geometry is identical with both collimation settings, the penumbra effect does not vary. Therefore, if the exposure parameters and section thickness are held constant, there is no significant radiation dose difference between the two settings. Although there is only a minor improvement in the quality of depiction of most abnormalities with an isotropic acquisition (Fig 13), there is no penalty for using the isotropic setting routinely for all cases. In other words, the ALARA principle does allow routine acquisition of isotropic data on 64-channel scanners.
View larger version (28K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 12a. Detector configurations and voxel dimensions at 64-channel multidetector CT. Because the width of the incident beam does not change between detector configurations, the concepts of narrow and wide collimation do not apply. (a) Left: Diagram shows the detector configuration for thin-section acquisitions. With each detector element used individually, 64 sections with a thickness of 0.625 mm each are acquired per gantry rotation, resulting in long-axis coverage of 40 mm. Right: Diagram shows that reconstructed voxels in this mode are isotropic, with approximately equal length in each dimension. (b) Left: Diagram shows the detector configuration for acquisition of thicker sections. Although the beam collimation does not change, the data acquisition system pairs the elements for the receipt of data. As a result, 32 sections with a thickness of 1.25 mm each are acquired per gantry rotation, while long-axis coverage remains constant. Right: Diagram shows that reconstructed voxels in this mode are anisotropic, approximately twice as large in the longitudinal plane as in the transverse plane.
|
|
View larger version (27K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 12b. Detector configurations and voxel dimensions at 64-channel multidetector CT. Because the width of the incident beam does not change between detector configurations, the concepts of narrow and wide collimation do not apply. (a) Left: Diagram shows the detector configuration for thin-section acquisitions. With each detector element used individually, 64 sections with a thickness of 0.625 mm each are acquired per gantry rotation, resulting in long-axis coverage of 40 mm. Right: Diagram shows that reconstructed voxels in this mode are isotropic, with approximately equal length in each dimension. (b) Left: Diagram shows the detector configuration for acquisition of thicker sections. Although the beam collimation does not change, the data acquisition system pairs the elements for the receipt of data. As a result, 32 sections with a thickness of 1.25 mm each are acquired per gantry rotation, while long-axis coverage remains constant. Right: Diagram shows that reconstructed voxels in this mode are anisotropic, approximately twice as large in the longitudinal plane as in the transverse plane.
|
|
View larger version (139K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 13a. Volume-rendered images from 64-channel multidetector CT. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Image reconstructed from anisotropic data (section thickness, 1.25 mm; increment, 0.625 mm) clearly depicts a peripheral aneurysm of the left renal artery. (b) Image reconstructed from isotropic data (section thickness, 0.625 mm; increment, 0.3 mm) provides sharper definition of vessel margins and allows visualization of small lumbar and mesenteric vessel branches.
|
|
View larger version (148K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 13b. Volume-rendered images from 64-channel multidetector CT. The first row of data on the images is the detector configuration, the middle row is the volume CT dose index, and the bottom row is the scanning time. (a) Image reconstructed from anisotropic data (section thickness, 1.25 mm; increment, 0.625 mm) clearly depicts a peripheral aneurysm of the left renal artery. (b) Image reconstructed from isotropic data (section thickness, 0.625 mm; increment, 0.3 mm) provides sharper definition of vessel margins and allows visualization of small lumbar and mesenteric vessel branches.
|
|
|
Conclusions
|
|---|
With increasing clinical workloads, it is difficult to keep pace with the implications of recent technologic advances. Despite our best intentions, many of us are ill prepared to balance the rather nebulous pros and cons that weigh in decisions concerning the optimization of image quality in multi-detector CT and the minimization of the radiation dose to the patient according to the ALARA principle. We hope this article will help generate discussion about the relative merits of isotropic and near-isotropic acquisitions. Fortunately, advances in CT scanner technology have led to a decrease in the negative trade-offs required. Nevertheless, changes in detector configuration on four- and 16-channel scanners (for a different reason with each scanner) may result in increases of approximately 100% in the radiation dose to the patient. With the 40-channel scanner platform, the difference in radiation dose between anisotropic and isotropic acquisitions is reduced to approximately 15%–20%. With the 64-channel scanner platform, there is essentially no trade-off, and routine acquisitions of isotropic data are justifiable. Ultimately, only quantitative analyses of the clinical utility of isotropic acquisitions can provide the information required to make informed decisions that balance the desire for high-quality images against the responsible use of ionizing radiation. Although some quantitative data are available (16–18), numerous variables affect image quality and radiation dose to the patient, and considerable further study is needed.
Top
Abstract
Introduction
