| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorial |
1 From the Department of Radiology, Boston University Medical Center, 88 E Newton St, Boston, MA 02118 (J.A.S.); Department of Radiology, Brigham and Women's Hospital, Boston, Mass (M.A.B.); and Department of Radiology, San Francisco Veterans Administration Medical Center, San Francisco, Calif (J.Y.). Received January 3, 2005; accepted January 10. Address correspondence to J.A.S. (e-mail: Jorge.Soto{at}bmc.org).
Although computed tomographic (CT) colonography is increasingly being recognized as a legitimate method of detecting colorectal neoplasia (1), an important condition for the broad acceptance of this examination will be the assurance of high-quality standards of performance and interpretation. As the technique continues to evolve and improve, its use is shifting from highly specialized academic centers to community hospitals and nonacademic radiology practices. Many radiologists are experiencing pressure from clinical colleagues to offer CT colonography as part of the routine services provided in their practices. However, CT colonography is not just an extension of CT performed for other indications; therefore, as with the training required before any new technique is applied, radiologists who are currently in practice should not be expected to be able to perform CT colonography or interpret the resultant findings without additional training.
There is emerging consensus that one of the most important factors affecting the diagnostic performance of CT colonography is the training and experience of the radiologist interpreting the examination (1). In a survey conducted among the majority of the members of the Working Group in Virtual Colonoscopy, an ad hoc group of investigators active in the area of CT colonography, the 18 respondents unanimously concurred that specific training is needed to interpret CT colonographic findings (2). The same group suggested that the most advisable method of training readers is an intensive hands-on course conducted at reading workstations with the supervision of experienced radiologists and by using exclusively cases with optical colonoscopy proof. During this training course, each reader should independently interpret between 40 and 50 cases that include examples of normal CT colonographic findings and a wide variety of lesion sizes and morphologic features, as can be expected in actual clinical practice. The premise is that this experience will allow the reader to become familiar with the various methods available for the primary interpretation of CT colonographic data sets while teaching him or her to recognize the technical variables and potential interpretation pitfalls that may affect the diagnostic performance of the examination. Formal lectures in which data supporting the use of CT colonography for colon screening are reviewed and that include discussions addressing patient preparation, technical acquisition parameters, interpretation techniques, and issues not related to interpretation are also advisable.
There is also emerging consensus that there is a variable learning curve associated with interpreting CT colonographic findings. This implies that interpretation performance improves as the number of interpreted cases increases. Although this may be true as a general statement, it is very likely that the specific characteristics of the curve vary between observers. Evidently, every observer will eventually reach an individual plateau of maximal performance. For example, in the multicenter trial conducted by Pickhardt et al (3,4), there was no significant difference in the diagnostic performance of the six readers when the four quartiles were analyzed separately. Thus, in that trial, every reader achieved a plateau of maximal performance after enrollment of 25% (approximately 50 patients per reader) of the patients.
Interpretation of CT colonographic findings entails a meticulous evaluation of the complete luminal surface of the colonic wall. Since this search pattern may be unfamiliar to most inexperienced observers, it is possible that some individuals, irrespective of how many cases they evaluate during the training process, may not ever reach the desired plateau of maximal performance. In other words, there is no guarantee that a given reader will achieve an adequate performance level by completing training sessions (5). Even if the desired performance level is to be reached, the slope of the curve may vary considerably between readers—from very steep (rapid improvement in performance with increasing training) to very shallow (slow improvement with increasing training). Moreover, the slope of the curve may vary according to the method used for primary interpretation (two-dimensional vs three-dimensional), even for the same individual. The starting point of the curve may also vary between observers. Readers with little or no experience in abdominal CT may require more specific training in CT colonography before they reach the optimal performance goals (6) compared with readers who have extensive experience in cross-sectional imaging of the abdomen and bowels.
Unfortunately, the lack of standards and uniformity in reader training and prior experience has plagued the prospective CT colonography clinical trials performed to date. The studies that have tended to yield good results have been single-center investigations performed by highly trained observers, many of whom are given credit for developing the technique (7–12). The results of larger multicenter trials conducted with patients who had an average or near-average risk of developing colorectal carcinoma have been more variable (3,13). These discrepancies have generated doubts about the true diagnostic capabilities of CT colonography, and, thus, the technique is now the target of careful scrutiny (14). Many factors are probably independent contributors to these discrepant results. These factors include hardware and software considerations, the quality of the CT data sets (specifically colonic distention and preparation), the use of stool and fluid tagging, the methods used for primary interpretation of the findings (two-dimensional vs three-dimensional images), the validity of the reference standard used, and, especially, the experience of the observers involved in data interpretation.
For their multicenter trial, Cotton et al (13) enrolled patients at nine centers, each with one or two radiologists who were required to have the prior experience of having interpreted the results of only 10 CT colonographic examinations each. One of the centers recruited 184 (29.9%) of the 615 patients enrolled and achieved a mean sensitivity of 82% for the detection of colonic adenomas 6 mm or larger. The mean sensitivity achieved by the remaining eight centers was 24%. What can this wide discrepancy in performance within the same trial be attributed to? In letters to the Editor published in the Journal of the American Medical Association (15–18), numerous researchers identified insufficient reader training and/or experience as one of the key factors responsible for the low sensitivity at many of the centers that participated in the Cotton et al study. At the end of the trial, each of the remaining eight centers had recruited an average of 54 patients. Since each center had one or two participating radiologists, it is probable that a given radiologist involved in this multicenter trial may have interpreted the CT colonographic findings of fewer than 27 patients during the recruitment period.
In contrast, each of the six radiologists who participated in the multicenter trial of Pickhardt et al (3) had interpreted the findings of at least 25 CT colonographic examinations (two radiologists may have interpreted the findings of at least 100 examinations) before the enrollment period began. At the end of that study, the average number of CT colonographic cases reviewed by the six radiologists exceeded 200. As noted earlier, other variables that were suggested as independent contributors to the wide discrepancies in the diagnostic performance of CT colonography in the studies of Pickhardt et al (3) and Cotton et al (13) included the use of stool and fluid tagging, the different hardware used for data acquisition, the different methods used for primary interpretation, and variations in the quality of the individual data sets.
Two large multicenter trials that are currently underway have much stricter requirements for prior reader experience. For the United Kingdom's National Health Service Special Interest Group in Gastrointestinal and Abdominal Radiology, or SIGGAR, 1 trial, all readers were tested before they were allowed to participate and additional training was offered if necessary (17). For the American College of Radiology Imaging Network II, or ACRIN II, trial, each reader had to demonstrate proficiency by taking a preenrollment certification examination involving the interpretation of 20 cases and demonstrate proficiency in CT colonogram interpretation by reaching a preset performance threshold. Those investigators who had interpreted fewer than 500 CT colonography cases with endoscopic correlation attended an intensive CT colonography training course before they were tested (J.G. Fletcher, written communication, December 20, 2004). Use of this prudent approach will probably eliminate (or at least greatly minimize) variation in reader experience as a factor affecting the ultimate results of such studies. The issue may then be whether or not these performance results can be extrapolated to less experienced readers.
Although we know that adequate training in CT colonography is necessary, we do not know how much training is sufficient to ensure adequate performance. We believe that CT colonography training should be incorporated into the curricula of diagnostic radiology residency and abdominal imaging fellowship programs. In the meantime, practicing radiologists must take specific training courses and/or participate in minifellowships before interpreting CT colonographic findings. Performing continued consistent interpretations after undergoing training sessions is probably important for sustaining adequate reader performance.
 |
References |
|---|
 TOP References |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. A. Taylor, J. Brittenden, J. Lenton, H. Lambie, A. Goldstone, P. N. Wylie, D. Tolan, D. Burling, L. Honeyfield, P. Bassett, et al. Influence of Computer-Aided Detection False-Positives on Reader Performance and Diagnostic Confidence for CT Colonography Am. J. Roentgenol., June 1, 2009; 192(6): 1682 - 1689. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S A TAYLOR, D BURLING, M RODDIE, L HONEYFIELD, J MCQUILLAN, P BASSETT, and S HALLIGAN Computer-aided detection for CT colonography: incremental benefit of observer training Br. J. Radiol., March 1, 2008; 81(963): 180 - 186. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. E. Baker, L. Bogoni, N. A. Obuchowski, C. Dass, R. M. Kendzierski, E. M. Remer, D. M. Einstein, P. Cathier, A. Jerebko, S. Lakare, et al. Computer-aided Detection of Colorectal Polyps: Can It Improve Sensitivity of Less-Experienced Readers? Preliminary Findings Radiology, October 1, 2007; 245(1): 140 - 149. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. S. Lee, S. H. Park, E. K. Choi, S. Y. Kim, M.-J. Kim, K. H. Lee, and Y. H. Kim Colorectal Polyps on Portal Phase Contrast-Enhanced CT Colonography: Lesion Attenuation and Distinction from Tagged Feces Am. J. Roentgenol., July 1, 2007; 189(1): 35 - 40. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. S. WERDERMAN CT Colonography Radiol. Technol., March 1, 2007; 78(4): 309CT - 324CT. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
