Published online before print June 23, 2003, 10.1148/radiol.2282020295
(Radiology 2003;228:569-575.)
© RSNA, 2003
Circular Tomosynthesis: Potential in Imaging of Breast and Upper Cervical Spine—Preliminary Phantom and in Vitro Study1
Grant M. Stevens, PhD,
Robyn L. Birdwell, MD,
Christopher F. Beaulieu, MD, PhD,
Debra M. Ikeda, MD and
Norbert J. Pelc, ScD
1 From the Departments of Radiology (G.M.S., R.L.B., C.F.B., D.M.I., N.J.P.) and Applied Physics (G.M.S.), Lucas MRSI Center, Stanford University, Calif; and GE Medical Systems Lunar, Mailstop 215, 726 Heartland Trail, Madison, WI 53717 (G.M.S). From the 2000 RSNA scientific assembly. Received March 12, 2002; revision requested June 3; final revision received November 22; accepted December 16. Supported in part by GE Medical Systems and the Lucas Foundation. Address correspondence to G.M.S. (e-mail: Grant.Stevens@med.ge.com).
 |
ABSTRACT
|
|---|
Phantom and in vitro studies were performed to evaluate the potential application of digital circular tomosynthesis in imaging of the breast and upper cervical spine. A prototype volumetric x-ray system was used to image a mammographic phantom, a fresh mastectomy specimen, and a head phantom containing the upper cervical spine. Results show that breast tissue visualization is improved by the ability to produce sectional images that blur overlying structures and yield three-dimensional information about calcification clusters. In upper cervical spine imaging, digital circular tomosynthesis effectively blurs overlying jaw and skull structures so that C1 and C2 can be visualized in a standard anteroposterior view.
© RSNA, 2003
Index terms: Images, processing • Phantoms • Radiography, digital • Radiography, technology • Test objects
|
INTRODUCTION
|
|---|
Interest in tomosynthesis and its clinical applications has been revived with recent advances in digital x-ray detector technology. Other authors have explored its use in mammography (1,2), dental imaging (3,4), angiography (5,6), and the detection of lung nodules (7), particularly with linear motion tomography. In this report, we focus on the potential application of circular motion tomosynthesis in which a reconstruction algorithm for enhanced performance (8) is used for mammography and imaging of the upper cervical spine.
The application of tomosynthesis to mammography is motivated by the fact that not all breast cancers are detected by using conventional radiography (9,10), and the cancers that are missed tend to be in radiographically dense breasts with overlying structures (10,11). From a diagnostic standpoint, calcifications on mammograms can be partly characterized as benign or malignant on the basis of their spatial distribution (12,13), as well as the form of the individual calcification. Therefore, tomosynthesis, in which overlying structures are removed from images of the area of interest, provides information regarding the relative three-dimensional (3D) positions of calcifications and could aid in screening for initial detection of cancers.
Imaging the upper cervical spine represents another area where tomosynthesis may be beneficial. For evaluating patients for cervical spine trauma, the American College of Radiology (ACR) currently recommends three views for radiographic screening: anteroposterior (AP), lateral, and open-mouth AP (14). The open-mouth view is needed because the mandible and teeth are superimposed over the cranial vertebrae in standard AP images. Imaging C1 and C2 through the open jaw represents an attempt to limit the effect of overlying structures. However, the highest-risk patients are often intubated (adding additional overlying structures that may not be able to be moved) or unable to comply with the imaging requirements owing to their injuries, and are therefore difficult to image with the open-mouth view (15,16). In fact, the most common location of missed fractures at the radiographic screening stage is the C1–C2 level (17,18). Conventional tomography (19,20) and computed tomography (CT) (16,17,21) have been performed to replace the open-mouth view for such patients. Conventional tomography, and hence tomosynthesis, may be preferable to CT because approximately 10% of cervical spine fractures are located at the dens (22,23), which conventional tomography may depict better than CT (15,18,24–26). Also, many fractures lie in the transverse plane, where they can be missed at CT. Tomosynthesis could potentially replace acquisition of the standard and open-mouth AP views by blurring the overlying structures in images obtained in patients who cannot be adequately imaged with radiography.
A tabletop system was previously developed to study the fundamentals of this type of imaging (27). Although this system (described below) is limited to use for the in vitro evaluation of tomosynthesis, data acquired with this system can generate insight into the potential for tomosynthesis to improve diagnosis. The goal of this study was to assess the potential application of digital circular tomosynthesis in imaging of the breast and upper cervical spine.
|
Materials and Methods
|
|---|
In this study, two phantoms and a specimen were used in the evaluation of the potential use of circular tomosynthesis: the ACR Mammography Quality Control phantom (28), a head phantom (Radiology Support Devices, Long Beach, Calif), and a fresh mastectomy specimen. The use of the fresh mastectomy specimen was approved by the institutional review board of Stanford University. The patient had signed a hospital consent form indicating consent for research-related activities. Our institutional review board granted a waiver of additional informed consent.
Data were collected by using a test-bed volumetric tomography system (27), as shown in Figure 1. This system consists of an x-ray tube (0.6-mm focal spot) and a 20 x 20-cm digital flat-panel detector with 200-µm pixels (GE Medical Systems, Milwaukee, Wis) mounted on a U arm (LU/A; GE Medical Systems). The object to be imaged is placed on a positioning stage, and views are acquired while the object is rotated through 360° while the tube and detector are kept fixed. For this system, the tomographic angle is defined as twice the angle between the central x ray and the axis of rotation of the positioning stage (Fig 1) and is controlled by the position of the U arm. No attempt was made to minimize radiation dose in this study, because the goals were qualitative rather than quantitative. Comparison radiographs of the phantoms were also acquired by using the same system.
View larger version (29K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Prototype digital volumetric tomography system. In circular tomosynthesis mode, the x-ray source and detector are rotated from vertical by  (half the tomographic angle) and locked in place. The object being imaged is placed on an x-ray-transparent positioning stage and rotated between each exposure through a total of 360°. The planes reconstructed for these experiments are parallel to the plane depicted here.
|
|
The tomosynthesis images were reconstructed by using filtered back projection with a blurring disk filter (8). In addition, both the radiographic and tomosynthesis images of the mastectomy specimen and the head phantom were processed after image reconstruction by using unsharp masking. The unsharp masking was performed by subtracting 75% of the pixel intensities in an image that had been smoothed (by using a 201 x 201-pixel Gaussian kernel, with a 52-pixel full width at half maximum) from the original radiographic and reconstructed image.
Breast Imaging and Image Review
Tomosynthesis data were collected by using the ACR phantom so that we could evaluate the ability of tomosynthesis to yield 3D information about the calcification clusters. Because all of the calcifications in each cluster were approximately co-planar, the phantom was tilted by approximately 20° with respect to horizontal (ie, the positioning stage) during image acquisition. This allowed the calcifications to lie on multiple planes parallel to the positioning stage; the orientation of these planes corresponded to the natural reconstruction planes of the system. A data set with 100 projection images and a 50° tomographic angle was obtained. Although it would be difficult for a clinical system to acquire 100 views in an acceptably short imaging time, this number of views was selected to demonstrate the ultimate potential of this type of imaging. The 100-projection data were additionally subsampled to reconstruct tomosynthesis images with 10 and 20 views. An x-ray technique of 40 kVp and 160 mA was used, with a 200-msec exposure time per view. The selection of this technique was limited by the x-ray tube and generator capabilities, as discussed below. So that we could compare tomosynthesis images with radiographs at equivalent total exposure, 100 projection radiographs of the ACR phantom were obtained by using the same x-ray technique.
To demonstrate the potential benefits of tomosynthesis compared with mammography in imaging human tissue, we imaged a fresh mastectomy specimen. The specimen was imaged (while wrapped in a towel) immediately after surgical excision, before slicing for pathologic examination. A 50° tomographic angle and 100 projection images were used for the tomosynthesis data set. As with the ACR phantom, the 100-projection data were subsampled to reconstruct tomosynthesis images with 10 and 20 views. An x-ray technique of 40 kVp and 160 mA was used, with a 200-msec exposure time per view. Because of the limited time allowed for handling the specimen before it was sliced for pathologic examination, only 10 radiographs could be acquired for comparison with the tomosynthesis images.
The images of the ACR phantom were independently reviewed by three of the authors (G.M.S., R.L.B., and N.J.P.). The images were qualitatively analyzed to determine the ability of tomosynthesis to blur representative mammographic structures outside the plane of interest. The mastectomy specimen images were independently reviewed by three authors (G.M.S., R.L.B., and N.J.P.) who qualitatively evaluated the blurring of off-plane tissues and lack of blurring of features in the focal plane. For both the phantom and specimen images, off-plane blurring was judged in terms of whether or not it was adequate and whether one reconstruction method was superior in this regard. There were no disagreements among the observers.
Cervical Spine Imaging and Image Review
To examine the potential usefulness of tomosynthesis in replacing the standard and open-mouth AP radiographic views, we imaged an anthropomorphic phantom consisting of plastic-encased bone-equivalent plastic in the shape of the skull and cervical spine. The mouth of the phantom could not be manipulated to mimic that of a patient in an open-mouth AP view, as is discussed below. Circular tomosynthesis data sets of the head phantom were acquired with tomographic angles of 10°, 20°, and 50°. Each data set contained 100 projection images. Additionally, the 100-projection data sets were subsampled to reconstruct tomosynthesis images with 20 and 50 views. An x-ray technique of 80 kVp and 160 mA was used, with an 80-msec exposure time per view. So that we could compare tomosynthesis images with radiographs at equivalent signal-to-noise ratio, 100 radiographs of the head phantom were obtained by using the same x-ray technique.
The images of the head phantom were reviewed by three of the authors (G.M.S., C.F.B., and N.J.P.) to assess the application of tomosynthesis to imaging of the cervical spine. We compared the ability of tomosynthesis to blur the jaw and skull structures lying outside the plane of interest with that of radiography by comparing the tomosynthesis images with a radiograph of the same phantom, and we qualitatively evaluated the tomosynthesis images in terms of the blurring of off-plane structures and the lack of blurring of features in the focal plane. Off-plane blurring was assessed in terms of whether or not it was adequate for cervical spine imaging and whether one reconstruction method was superior in this regard. There were no disagreements among the observers.
|
Results
|
|---|
Breast Imaging
A single-projection mammogram provided only planar information about the calcifications in the tilted ACR phantom, as shown in Figure 2. In comparison, circular tomosynthesis images of this calcification group yielded 3D information. On the tomosynthesis images, the calcifications that lay outside the reconstruction plane were blurred, as shown in Figure 3.
View larger version (113K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2. Digital radiograph of the largest calcification group in the tilted ACR phantom. This image provides no depth information regarding the 3D distribution of the calcifications.
|
|
View larger version (74K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3. Tomosynthesis images (50° tomographic angle, 100 views) of the largest calcification group in the ACR mammography phantom. Each image is separated from its neighbor (left to right and top to bottom) by 1 mm. The images show that the calcifications (arrows) in the lower left corner of the calcification group are closer to the x-ray tube than those in the upper right corner.
|
|
The difficulty of discerning structures in a radiographic image of the breast due to the presence of overlying structures is evident in a radiograph of the mastectomy specimen (Fig 4). The use of tomosynthesis improved visualization of the underlying tissue detail by blurring the overlying structures. Figure 5 shows a single section from the tomosynthesis reconstructions of the mastectomy specimen focused just below the skin flap. The skin flap, nipple, and calcification that are present in the radiograph (Fig 4) are effectively removed by the blurring in the 100-view tomosynthesis image (Fig 5a). This allows better visualization of the tissue detail directly below the skin and the nipple and areolar complex.
View larger version (174K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 4. Radiograph of a fresh mastectomy specimen, with the location of the nipple, skin flap, and a calcification noted. During the imaging process, the tissue specimen was wrapped in a towel, which is evident as an angled band across the bottom of the image and as the structured pattern over the entire image.
|
|
View larger version (137K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5a. Tomosynthesis images (50° tomographic angle) of the mastectomy specimen with (a) 100, (b) 20, and (c) 10 views show individual linear Cooper ligaments and glandular elements below the nipple and areolar complex with greater clarity than the projection radiograph in Figure 4. In a, the overlying nipple and skin and the calcification are blurred. Use of fewer views for reconstruction produces a star-shaped artifact (arrow in b and c) owing to the discrete blurring of the calcification. Furthermore, c shows the towel in which the specimen was placed.
|
|
View larger version (144K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5b. Tomosynthesis images (50° tomographic angle) of the mastectomy specimen with (a) 100, (b) 20, and (c) 10 views show individual linear Cooper ligaments and glandular elements below the nipple and areolar complex with greater clarity than the projection radiograph in Figure 4. In a, the overlying nipple and skin and the calcification are blurred. Use of fewer views for reconstruction produces a star-shaped artifact (arrow in b and c) owing to the discrete blurring of the calcification. Furthermore, c shows the towel in which the specimen was placed.
|
|
View larger version (151K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 5c. Tomosynthesis images (50° tomographic angle) of the mastectomy specimen with (a) 100, (b) 20, and (c) 10 views show individual linear Cooper ligaments and glandular elements below the nipple and areolar complex with greater clarity than the projection radiograph in Figure 4. In a, the overlying nipple and skin and the calcification are blurred. Use of fewer views for reconstruction produces a star-shaped artifact (arrow in b and c) owing to the discrete blurring of the calcification. Furthermore, c shows the towel in which the specimen was placed.
|
|
Tomosynthesis images of the mastectomy specimen reconstructed with 20 and 10 views are shown in Figure 5b and Figure 5c, respectively. The image with 20 views (Fig 5b) shows that the large calcification has been blurred into discrete points, producing a star-shaped artifact owing to reconstruction with an insufficient number of views for a 50° tomographic angle. This artifact is more evident in Figure 5c, for which only 10 views were used. In addition, the skin flap is less adequately blurred in the 20-view reconstruction. Even so, the region below the skin flap can be seen quite well. In the 10-view reconstruction, the skin flap is somewhat blurred but is still visible, and the structure of the towel covering the tissue specimen can still be seen.
Cervical Spine Imaging
The entire upper cervical spine is clearly depicted in a lateral-view radiograph of the head phantom (Fig 6a). An AP view of the phantom, however, does not adequately display the upper cervical spine. As shown in Figure 6b, C1 and C2 are obscured because of the interference of the overlying jaw structures.
View larger version (119K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6a. (a) Lateral radiograph of the head phantom demonstrates the excellent visualization of the top of the cervical spine with this view, while (b) AP radiograph highlights the difficulty in visualizing the upper cervical spine with this view.
|
|
View larger version (105K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 6b. (a) Lateral radiograph of the head phantom demonstrates the excellent visualization of the top of the cervical spine with this view, while (b) AP radiograph highlights the difficulty in visualizing the upper cervical spine with this view.
|
|
Tomosynthesis images obtained with 100 views by using a 50° tomographic angle (Fig 7a) clearly showed C1 and C2. The use of tomosynthesis allowed better visualization of the cervical spine owing to the blurring of the anatomic structures above and below the visualized plane. If the number of views is not sufficient, however, artifacts arise (8). As Figure 7b demonstrates, these artifacts are present in reconstructions of a 50°-tomographic-angle data set with 50 views and are even more apparent when the number of views is further decreased to 20 (Fig 7c).
View larger version (105K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7a. Tomosynthesis images (50° tomographic angle) of the head phantom, with (a) 100, (b) 50, and (c) 20 views. As the number of views is reduced, the artifacts in the reconstructed image become more prevalent; substantial artifacts are seen in c.
|
|
View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7b. Tomosynthesis images (50° tomographic angle) of the head phantom, with (a) 100, (b) 50, and (c) 20 views. As the number of views is reduced, the artifacts in the reconstructed image become more prevalent; substantial artifacts are seen in c.
|
|
View larger version (124K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 7c. Tomosynthesis images (50° tomographic angle) of the head phantom, with (a) 100, (b) 50, and (c) 20 views. As the number of views is reduced, the artifacts in the reconstructed image become more prevalent; substantial artifacts are seen in c.
|
|
To reduce the number of views required to adequately image this region, the tomographic angle can be decreased. Figure 8 shows images reconstructed from the 20°-tomographic-angle data set. As with the 50°-tomographic-angle data set, 100 views are sufficient to clearly depict C1 and C2 (Fig 8a). For this smaller tomographic angle, in fact, 50 views are sufficient, as shown in Figure 8b. This reduction in the required number of views comes at the cost of a larger section thickness, as is evident by the loss of some detail in the images in Figure 8 relative to those in Figure 7. This compromise, however, may be necessary if the imaging protocol requires a reduction in the number of views. When the number of views is reduced to 20 with a 20° tomographic angle, artifacts are again visible (Fig 8c). The tomographic angle cannot be made too small, however, owing to the resulting increase in section thickness. As is evident in Figure 9, use of a 10° tomographic angle yields images in which out-of-plane structures are insufficiently blurred owing to the large section thickness.
View larger version (104K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8a. Tomosynthesis images (20° tomographic angle) of the head phantom, with (a) 100, (b) 50, and (c) 20 views. The reduction in the amount of artifacts in the images obtained with fewer views—as compared with the images in Figure 7—comes at the cost of a larger section thickness.
|
|
View larger version (108K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8b. Tomosynthesis images (20° tomographic angle) of the head phantom, with (a) 100, (b) 50, and (c) 20 views. The reduction in the amount of artifacts in the images obtained with fewer views—as compared with the images in Figure 7—comes at the cost of a larger section thickness.
|
|
View larger version (119K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 8c. Tomosynthesis images (20° tomographic angle) of the head phantom, with (a) 100, (b) 50, and (c) 20 views. The reduction in the amount of artifacts in the images obtained with fewer views—as compared with the images in Figure 7—comes at the cost of a larger section thickness.
|
|
View larger version (93K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 9. Tomosynthesis image (10° tomographic angle, 100 views) of the head phantom. At this tomographic angle, the section thickness is too large to result in adequate blurring of overlying structures.
|
|
|
Discussion
|
|---|
The tomosynthesis images in this article were acquired by using a circular motion of the x-ray tube and detector. Alternatively, tomosynthesis (as well as conventional motion tomographic) images can be acquired by using other motions. The type of motion used during data acquisition dictates the type of blurring of off-focal-plane objects in the image. Simple linear motion blurs objects in one dimension only, which leads to linear streak artifacts caused by high-contrast off-focal-plane objects (29). Conventional use of circular motion of the source and detector results in out-of-plane points being blurred into rings, while more complex motions, such as hypocycloidal motion, result in more (visually) optimal blurring functions (30). The drawback to the use of more complex motions, however, is the need for more costly equipment and potentially longer exposures (31). In addition, although this report describes images obtained by using circular motion, the images were reconstructed by using a technique that can produce images with blurring that is equivalent to that seen in images obtained with more complex motions (8).
The work reported herein was performed with a prototype system that simplified the system requirements for the imaging gantry and detector. To collect clinical data, one would have to use a detector that could work with an entrance dose low enough to allow sufficient signal-to-noise ratio at tomosynthesis doses and that could output data fast enough to quickly image a region. The digital flat-panel detectors currently being developed for fluoroscopic use could be potential targets for clinical application. The high conversion factor of such detectors would allow for a low dose for each view, and fluoroscopic imaging rates could allow for an acquisition within a breath hold.
Breast Imaging
The images presented in this article demonstrate the feasibility of circular tomosynthesis for revealing the spatial orientation of calcifications and for visualizing breast tissue structures below the skin. An optimized mammography system would produce even better images for two main reasons. First, the prototype system we used was not designed to operate in mammographic conditions (ie, at 26 keV with a molybdenum target and filter). The lowest peak voltage of the prototype system is 40 keV, the x-ray tube has a tungsten target, and the x-ray source has at least 2.5-mm aluminum equivalent filtration at 80 keV (Ultranet SA operator manual, GE Medical Systems). The findings of our initial computer simulations showed that these factors result in an x-ray beam with an average energy of approximately 27 keV (compared with <20 keV with a typical clinical mammography system), and, more importantly, that all x rays with energies of less than 20 keV are effectively filtered from the beam. Second, conventional screen-film systems have spatial resolution up to 20 lp/mm (ie, 25-µm pixels), although limits in contrast- and signal-to-noise ratios at mammography mean that this resolution capability is not fully utilized (32). Mammographic digital detectors have pixels on the order of 50–100 µm (32,33), versus the 200-µm pixels of the general radiography detector we used. The combined effect of these differences in the parameters of the prototype system was the generation of mammograms with reduced spatial resolution and reduced contrast-to-noise ratio relative to images obtained with mammography-specific hardware.
Although the system we used was not optimized for mammographic applications, the tomosynthesis images could still be judged relative to radiographs obtained with the same x-ray system. Thus, we were able to demonstrate the improved performance of tomosynthesis relative to projection radiography but not the full measure of its capabilities. Application of the same imaging techniques with hardware optimized for mammography is expected to yield similar relative improvements.
As noted above, use of a smaller tomographic angle increases the section thickness of the reconstructed image. This limits the blurring of nearby out-of-plane objects and decreases the observer’s ability to both visualize objects and determine their 3D location. It should be noted, however, that at times it may be desirable to use smaller tomographic angles. In addition to the fact that geometric limitations (eg, positioning requirements) may dictate the use of small tomographic angles, use of a smaller tomographic angle is beneficial in situations (eg, when there are time constraints) in which fewer views are required owing to patient motion or in situations in which there are dose restrictions per view due to the noise floor of the detector. The smaller blur radius of out-of-plane structures resulting from use of a small tomographic angle reduces the discrete nature of the blurring. This lessens the chance that an observer will interpret discretely blurred structures as additional on-plane objects and improves visualization of objects in the focal plane at the cost of a decreased total extent of blurring.
Results of our experiments with the mastectomy specimen suggest that for a 50° tomographic angle, 20 views is close to the minimum number required to adequately image this amount of breast tissue. For tomosynthesis in vivo, where the thickness of breast tissue may be somewhat greater, the number of views may need to be higher.
Because our intent was to perform a qualitative analysis of circular tomosynthesis for mammography in this study, we made no attempt to reduce the radiation dose to the phantoms. Clearly this is an important consideration in addressing the applicability of this type of imaging to the clinical setting. Assuming a low noise detector, such as that discussed above, is used, the dose requirements and resulting image noise for tomosynthesis are the same as those for acquisition of a radiograph of the same anatomic region. The quantum noise in the detected x rays for each of the views comprising the tomosynthesis data set combines as a linear sum to equal that of a radiograph obtained with the same total dose. This allows for a direct comparison between radiographs and tomosynthesis images regardless of dose, assuming the same total dose is used for both acquisitions. Therefore, the amount of the dose delivered is not expected to limit the use of tomosynthesis, and results of our comparison of relative image quality can be extrapolated to use of clinically relevant doses.
Cervical Spine Imaging
The motivation for studying the potential use of tomosynthesis in imaging of the upper cervical spine was the goal of replacing the standard and open-mouth AP radiographs. Because the mouth of the phantom used in these experiments could not be manipulated, an open-mouth view could not be obtained. Therefore, tomosynthesis images were compared with standard AP radiographs of the phantom. Results of this comparison are most appropriate for the subset of patients who have sustained trauma and in whom the open-mouth view cannot be obtained, but the results can be extrapolated to suggest the general improvement that might be obtained by using tomosynthesis in imaging this region.
The trade-off in tomosynthesis imaging between section thickness and number of views was evident in our study in the images of the cervical spine. The ideal approach to tomosynthesis imaging would be to first determine the desired section thickness. This dictates the tomographic angle needed and therefore—along with the overall thickness of the object—governs the number of views needed. Our results indicate that a 50° tomographic angle appears to be sufficient for visualizing the upper cervical spine. With our reconstruction algorithm, approximately 100 views were needed for a 50° tomographic angle. If a larger section thickness is acceptable, a 20° tomographic angle can be used with 50 views.
An alternative approach to tomosynthesis imaging is to determine the number of views that can be acquired given imaging constraints (eg, time restrictions from patient motion, dose restrictions of the detector). The tomographic angle can be selected to yield images with an acceptable amount of artifacts; the tomographic angle then determines the achievable section thickness.
It appears that the number of views required for visualizing the cervical spine is quite large because of the complex structures present in this region. This is certainly the case with our algorithm, as indicated in this report, and would also be the case with simple back projection (or a shift-and-add algorithm). It is possible, however, that this restriction could be lessened by use of an alternative reconstruction scheme. We have limited our evaluation of tomosynthesis reconstruction schemes to back projection, but other techniques such as iterative reconstruction should be examined if the current techniques prove too limiting for clinical application.
Overall, the application of tomosynthesis to imaging of the upper cervical spine appears quite promising. The visualization of C1 and C2 on tomosynthesis images is far superior to that on a standard AP radiograph. In addition, the flexibility in the choice of tomosynthesis imaging parameters based on the desired final images and realistic imaging conditions may be beneficial.
In conclusion, use of circular tomosynthesis in imaging of the breast and upper cervical spine appears promising. Results of mammographic experiments suggest that tomosynthesis can improve image quality compared with conventional radiography by removing overlying structures and providing limited 3D information. Images of the head phantom used in this study suggest that visualization of C1 and C2 is improved on tomosynthesis images compared with AP radiographs owing to the removal of the overlying jaw structures.
The techniques developed for the prototype imaging system used in this study can be modified to be used in a clinical setting. With a clinical system, the patient would most likely remain stationary while the x-ray tube and detector move around the patient. This can be accomplished by using, for example, a motion tomography gantry. The image reconstruction techniques presented in this report require only minor modifications to be used in the clinical setting. These technologic obstacles should be quickly overcome if the clinical demand for this type of imaging is present.
|
ACKNOWLEDGMENTS
|
|---|
The authors thank Rebecca Fahrig, PhD, of Stanford University for technical assistance and Teri Longacre, MD, of Stanford University for assistance with the mastectomy specimen.
|
FOOTNOTES
|
|---|
Abbreviations: ACR = American College of Radiology,
AP = anteroposterior,
3D = three-dimensional
Author contributions: Guarantor of integrity of entire study, N.J.P.; study concepts, N.J.P., G.M.S.; study design, all authors; literature research, G.M.S.; clinical studies, G.M.S., R.L.B.; experimental studies and data acquisition, G.M.S.; data analysis/interpretation, G.M.S., N.J.P., C.F.B., D.M.I.; manuscript preparation, G.M.S.; manuscript definition of intellectual content and editing, N.J.P., G.M.S.; manuscript revision/review and final version approval, N.J.P.
|
REFERENCES
|
|---|
- Niklason LT, Christian BT, Niklason LE, et al. Digital tomosynthesis in breast imaging. Radiology 1997; 205:399-406.[Abstract/Free Full Text]
- Webber RL, Underhill HR, Freimanis RI. A controlled evaluation of tuned-aperture computed tomography applied to digital spot mammography. J Digit Imaging 2000; 13:90-97.[Medline]
- Gröenhuis RAJ, Webber RL, Ruttimann UE. Computerized tomosynthesis of dental tissues. Oral Surg Oral Med Oral Pathol 1983; 56:206-214.[CrossRef][Medline]
- Webber RL, Messura JK. An in vivo comparison of diagnostic information obtained from tuned-aperture computed tomography and conventional dental radiographic imaging modalities. Oral Surg Oral Med Oral Pathol 1999; 88:239-247.
- Stiel GM, Stiel LSG, Nienaber CA. Potential of digital flashing tomosynthesis for angiocardiographic evaluation. J Digit Imaging 1992; 5:194-205.[Medline]
- Haaker P, Klotz E, Koppe R, Linde R, Möller H. A new digital tomosynthesis method with less artifacts for angiography. Med Phys 1985; 12:431-436.[CrossRef][Medline]
- Dobbins JT, III, Webber RL, Hames SM. Tomosynthesis for improved pulmonary nodule detection (abstr). Radiology 1998; 209(P):280.
- Stevens GM, Fahrig R, Pelc NJ. Filtered backprojection for modifying the impulse response of circular tomosynthesis. Med Phys 2001; 28:372-379.[CrossRef][Medline]
- Baines CJ, Miller AB, Wall C, et al. Sensitivity and specificity of first screen mammography in the Canadian National Breast Screening Study: a preliminary report from five centers. Radiology 1986; 160:295-298.[Abstract/Free Full Text]
- Bassett LW, Bunnell DH, Jahanshahi R, Gold RH, Arndt RD, Linsman J. Breast cancer detection: one versus two views. Radiology 1987; 165:95-97.[Abstract/Free Full Text]
- Bird RE, Wallace TW, Yankaskas BC. Analysis of cancers missed at screening mammography. Radiology 1992; 184:613-617.[Abstract/Free Full Text]
- Sickles EA. Breast calcifications: mammographic evaluation. Radiology 1986; 160:289-293.[Abstract/Free Full Text]
- Franceschi D, Crowe J, Zollinger R, et al. Biopsy of the breast for mammographically detected lesions. Surg Gynecol Obstet 1990; 171:449-455.[Medline]
- Keats TE, Dalinka MK, Alazraki N, et al. Cervical spine trauma In: ACR appropriateness criteria. Reston, Va: American College of Radiology, 2000; 243-246.
- Ehara S, El-Khoury GY, Clark CR. Radiologic evaluation of dens fracture: role of plain radiography and tomography. Spine 1992; 17:475-479.[Medline]
- Blacksin MF, Lee HJ. Frequency and significance of fractures of the upper cervical spine detected by CT in patients with severe neck trauma. AJR Am J Roentgenol 1995; 165:1201-1204.[Abstract/Free Full Text]
- Nuñez DB, Zuluaga A, Fuentes-Bernardo DA, Rivas LA, Becerra JL. Cervical spine trauma: how much more do we learn by routinely using helical CT? RadioGraphics 1996; 16:1307-1318.[Abstract]
- Clark CR, Igram CM, El-Khoury GY, Ehara S. Radiographic evaluation of cervical spine injuries. Spine 1988; 13:742-747.[CrossRef][Medline]
- Ridpath CA, Wilson AJ, Langer SG, Mann FA, Hunter JC. Cervical spine tomography with an angiographic C-arm. Radiology 1999; 211:882-885.[Abstract/Free Full Text]
- Vandemark RM, Fay ME, Porter FR, Johnson GA. Digital image-intensifier radiography at a level I trauma center. AJR Am J Roentgenol 1997; 168:944-946.[Free Full Text]
- Peh WCG, Cheng P, Chan FL. Direct coronal computed tomography of the upper cervical spine. Spine 1995; 20:972-974.[Medline]
- Lipson SJ. Fractures of the atlas associated with fractures of the odontoid process and transverse ligament ruptures. J Bone Joint Surg Am 1977; 59:940-943.[Abstract/Free Full Text]
- Miller MD, Gehweiler JA, Martinez S, Charlton OP, Daffner RH. Significant new observations on cervical spine trauma. AJR Am J Roentgenol 1978; 130:659-663.[Abstract]
- Baumgarten M, Mouradian W, Boger D, Watkins R. Computed axial tomography in C1-C2 trauma. Spine 1985; 10:187-192.[Medline]
- Harris JH, Jr, Mirvis SE. The radiology of acute cervical spine trauma Baltimore, Md: Williams & Wilkins, 1996; 180-211.
- Woodring JH, Lee C. The role and limitations of computed tomographic scanning in the evaluation of cervical trauma. J Trauma 1992; 33:698-708.[Medline]
- Stevens GM, Saunders R, Pelc NJ. Alignment of a volumetric tomography system. Med Phys 2001; 28:1472-1481.[CrossRef][Medline]
- Mammography quality control manual Reston, Va: American College of Radiology, 1999.
- Macovski A. Medical imaging systems Englewood Cliffs, NJ: Prentice Hall, 1983.
- Carter SJ, Martin JJ, Middlemiss JH, Ross FGM. Polytome tomography. Clin Radiol 1963; 14:405-413.[CrossRef][Medline]
- Stanton L. Conventional tomography. In: Taveras JM, Ferrucci JT, eds. Radiology: diagnosis-imaging-intervention. Philadelphia, Pa: Lippincott, 1988.
- Feig SA, Yaffe M. Digital mammography. RadioGraphics 1998; 18:893-901.[Medline]
- Vedantham S, Karellas A, Suryanarayanan S, et al. Full breast digital mammography with an amorphous silicon-based flat panel detector: physical characteristics of a clinical prototype. Med Phys 2000; 27:558-567.[CrossRef][Medline]
This article has been cited by other articles:
|
|
|
|
 
T Gomi, N Yokoi, and H Hirano
Evaluation of digital linear tomosynthesis imaging of the temporomandibular joint: initial clinical experience and evaluation
Dentomaxillofac. Radiol.,
December 1, 2007;
36(8):
514 - 521.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
