| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Science to Practice |
Department of Medical Imaging,
Penn Presbyterian Medical Center,
39th and Market Streets,
Philadelphia, PA 19104
seth.glick{at}uphs.upenn.edu
SUMMARY
In the study by Kim et al, primary two-dimensional (2D) search with a three-dimensional (3D) problem-solving technique was compared with a 3D virtual dissection technique, with use of two independent reviewers. The authors observed no significant difference (P > .05) in sensitivity or specificity between the two techniques for the detection of polyps 6 mm or larger, whereas the median review time with virtual dissection was significantly (P < .05) shorter. This may have implications for the application of computed tomographic (CT) colonography for screening the asymptomatic average-risk population.
THE SETTING
Screening for the detection of early cancer and potentially premalignant polyps has become widely accepted. Colonoscopy is considered an accurate screening method and has the capability for polypectomy, but it is an invasive procedure that requires sedation and is associated with substantial complications. CT colonography has shown promise as a noninvasive alternative, but it has not received endorsement owing to the range of reported performance (1,2). Some of the inconsistent results have been attributed to reader experience, the technology and protocol used for image acquisition, and variability in image display methodology. While most study investigators have used a primary 2D search approach, the best outcomes have occurred with primary 3D display. It has also been demonstrated not only that there is a steep learning curve but also that some individuals may possess more innate potential to develop the necessary skills (3). In addition, lengthy interpretation times may be problematic for busy radiology practices when the CT colonography volume increases substantially. In this issue of Radiology, Kim et al (4) evaluated the performance of two readers with regard to accuracy and time efficiency when they used both a conventional 2D display mode and a 3D virtual dissection (filet) perspective that is used essentially to open and flatten the mucosal surface as if examining a surgical specimen.
THE SCIENCE
The protocol of Kim et al (4) integrated the prone and supine acquisitions for synchronous viewing. The readers performed a blinded retrospective review of the data sets from 96 consecutive CT colonography examinations. All patients had subsequently undergone segmental unblinded colonoscopy, and accepted polyp-matching criteria were applied. With 3D display, the performance of both readers in terms of per-polyp sensitivity (77% and 69% with 3D vs 69% and 63% with 2D), per-patient sensitivity (77% and 73% with 3D vs 64% and 59% with 2D), and per-patient specificity (99% and 89% with 3D vs 91% and 89% with 2D) for the detection of polyps 6 mm in diameter or larger was superior but not significantly different (P > .05) from their performance with 2D polyp detection. However, for both readers, the 3D mode was associated with a significant decrease (P < .05) in median interpretation time (9.4 and 9.6 minutes with 3D vs 14.1 and 14.4 minutes with 2D).
|
Clinical use:
At present, it is unclear how CT colonography will be used for colorectal cancer screening. There are not only existing alternatives (eg, conventional optical colonoscopy) and developing potential options (eg, new endoscopic techniques, genetic stool testing) (5) but also evolving CT colonography technologies, acquisition protocols, and display modes. Additional factors, such as quality assurance requirements and potentially large procedure volumes in the context of the busy radiologic practice, represent important considerations. Kim et al (4) demonstrated greater time efficiency and comparable performance with 3D virtual dissection when the target lesion was a polyp 6 mm in diameter or larger. However, the sensitivities reported, although slightly higher than those in some studies, were inferior to the best-reported results (6). Whether the actual values Kim et al report are considered acceptable for screening is a matter that might be debated. In addition, Kim et al cite a previous study (7) of virtual dissection where the performance did not reach the level of performance in their present investigation. While possible explanations for the variance are offered, they remain speculative. Furthermore, the analysis of interpretation time did not include evaluation of the extracolonic structures that could have had an effect on the time difference.
Future opportunities and challenges:
The concentration and focus required to review CT colonography data are considerable, and there is a relationship between volume and fatigue (8). The relationship between caseload, display mode, accuracy, and interpretation time is important. Finally, there are parallel areas of research such as low-dose technique (for reduction of radiation exposure), prepless or minimal preparation approaches, and computer-aided detection (9). It is unclear how 3D virtual dissection display will interface or compete with these other innovations. For CT colonography to have an important role in colorectal cancer screening, it will need to demonstrate consistent acceptable accuracy and cost-effectiveness, appeal to patients, minimize patient risks, and be time efficient.
FOOTNOTES
See also the article by Kim et al in this issue.
References
Related Article
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
