| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Commentary |
1 Department of Radiology, Imaging Research, University of Pittsburgh, 3362 Fifth Ave., FARP Bldg., Rm. 223, Pittsburgh, PA 15213-3180.
Received May 16, 2007; accepted after revision June 26, 2007.
Address correspondence to D. Gur
(gurd{at}upmc.edu).
Keywords: breast cancer • breast cancer screening • diagnostic imaging • mammography • tomosynthesis
In this issue of the AJR, Dr. Poplack and colleagues [1] explore one possible intended use of tomosynthesis-based breast imaging—namely, to significantly reduce recall rates. The development of digital detectors for mammography has enabled this exciting approach to breast imaging; however, the role of tomosynthesis in routine clinical practice is yet to be determined. In this commentary, I review several issues related to the ultimate incorporation of tomosynthesis into screening and diagnostic breast imaging practices.
Poplack and colleagues [1] present an interesting article about an exciting topic that we will be reading about frequently in the near future—namely, tomosynthesis-based imaging of the breast. The recent development of digital detectors for mammography has enabled us to examine using tomosynthesis in a way that is quite practical and that may be relatively easily implemented on several digital systems being used in radiology in general and for breast imaging in particular [2-5].
Although the current interest is primarily in using tomosynthesis for breast imaging, the fact is that this approach is relevant to several procedures, such as chest imaging. Tomosynthesis-based imaging is of great interest in screening mammography and diagnostic mammography for several reasons. One important aspect explored in the article by Poplack et al. [1] is the possibility of reducing recall rates in screening mammography.
The preliminary assessment presented in the article [1] suggests that tomosynthesis-based breast imaging could result in a significant reduction in recall rates during screening. Unfortunately, the study design does not permit the authors of this preliminary study to accurately assess changes in recall rates. Poplack et al. were aware of the need for corrections for the different recall rates of the original interpreters and the study radiologists. However, in the Discussion section, they allude only briefly to one fundamental issue: The fact that in a screening environment in the United States there are approximately nine cases that were not recalled during screening for every case that was recalled, and in other countries the number is even greater. For example, in some countries in Europe the recall rate is approximately 3%; hence, there are more than 30 cases not recalled for every one that is. The mere fact that a review of 98 tomosynthesis examinations (with 99 findings) that had been actually recalled by the radiologists during the original mammography interpretations generated eight additional recalls of a "different site stemming from the tomosynthesis examination" should have been recognized as the primary limitation of the study.
As designed, a study of only recalled cases was bound to show a potential improvement in recall rates. If we take the numbers from the study and apply them to the U.S. screening population—that is, we assume that the "other recalled sites" are representative of cases that would be potentially recalled in negative cases and that approximately 8% of the cases interpreted as negative on mammography would be recalled during viewing of the tomosynthesis-generated images—the total number of actual recalled cases will increase by 32% (i.e., 60% of the actual recalled cases constituting 10% of the sampled population and 8% of the originally negative cases on mammography that are now recalled based on tomosynthesis and constituting 90% of the sampled population or a total of 6% + 7.2% = 13.2% recall rate).
Even if we assume that the tomosynthesis-generated recall rate would be significantly lower in the mammography-negative population than the assumed 8%, one must recognize that but a small added recall rate (e.g., 4%) in this group constituting 90% of the population will rapidly diminish the benefit of tomosynthesis in this regard. This scenario worsens as actual recall rate by mammography alone decreases. Therefore, all studies that attempt to address recall rates in the general screening population when tomosynthesis is an integral part of the procedure must include a substantial number of representative cases that had originally not been recalled during mammography. We assume that the study radiologists in this article [1] are actually well trained in interpreting tomosynthesis-generated sets of images.
Unlike simple image-processing routines, the appropriate, accurate, and efficient use of tomosynthesis will necessitate substantial training not only in the appearances of different abnormalities but also in the appearances of the widely varying normal tissue leading to negative findings. As evidenced from the study [1], even those who were familiar with the procedure and had substantial experience in viewing tomosynthesis images generated recall rates in "other sites" that were not negligible. The standards and requirements for such training have not yet been determined and are beyond the scope of this commentary.
It is clear to all involved in this area that visualization tools must be developed to allow an efficient optimal assessment of the multiple images generated by tomosynthesis. Currently, extensive work is being done in this area to address the viewing of both masses and microcalcification clusters. Efficiency-enhancing visualization tools will be of particular importance if tomosynthesis-generated images are to be routinely used in the screening environment. Other important issues, such as the need for two-view tomosynthesis procedures versus one view (e.g., mediolateral oblique only) and the optimal acquisition techniques, will eventually be addressed.
An important diagnostic application that may be considered is the role of tomosynthesis in diagnostic mammography, particularly for ruling out suspected abnormalities that were identified during screening (e.g., masses and architectural distortions) and determining whether a visualized mass is likely to be benign or malignant. Improved visualization of overlying tissues and mass margins may ultimately prove to be very valuable in this regard.
The recent recommendation to periodically image women who are at "high risk" for breast cancer with MRI may result in substantial operational and added costs of screening for early detection of breast cancer in the screening population as a whole. Tomosynthesis may ultimately play an important role in screening a fraction of women at high risk, but much work is needed in this regard before we can assess its utility in at least some segments of the high-risk population.
Despite the very positive initial reaction to tomosynthesis by most radiologists in the field, the utility, clinical relevance, and cost-effectiveness of tomosynthesis are unknown. It will take some time and a number of rigorously executed studies to find how best to incorporate tomosynthesis into routine clinical practice. We also must be cognizant of the fact that tomosynthesis may ultimately be considered by some as but one step forward toward using CT in clinical breast imaging and that the initial experience with CT of the breast is as favorable as that with tomosynthesis. However, tomosynthesis does have significant operational, cost, and ease-of-use advantages over CT that should not be ignored. Hence, the approach should neither be taken lightly nor be seen as a temporary advance that is likely to be soon replaced by a more advanced technology. The fact that tomosynthesis is relatively easily and inexpensively adapted from a procedure that is practiced every year on many millions of women is, by itself, important enough. Therefore, we should investigate all the possible benefits that could be gained with this approach to breast imaging in both the screening and the diagnostic environments.
In summary, in the digital era tomosynthesis may have great potential for imaging the breast and other organs. The initial reaction of radiologists and other health care professionals to this approach is very encouraging. We must make every effort to ensure that future studies are well designed with meticulous scientific method and large patient cohorts to withstand scientific scrutiny and the test of time. Otherwise, this exciting new technology may not become incorporated into routine clinical practice.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
