In computed tomographic (CT) colonography, also known as virtual colonoscopy, patients are typically scanned in two positions . By comparing the two CT scans, radiologists can maximize the visible region of colon for polyp detection and also identify mobile residual materials or normal anatomy that could imitate abnormalities in a single scan. However, current computer-aided detection (CADe) systems consider each scan as a separate study; therefore, the locations of CADe prompts do not necessarily match between the two position scans of the same patient. Thus, radiologists are expected to review all regions indicated by CADe prompts alike, regardless of whether the region was prompted on both scans or on only one.
This month’s issue of Academic Radiology contains a research study that investigated the effect of “dual-matching” true-positive CADe prompts on the polyp identification performance of a computer-assisted reader in CT colonography . Four radiologists reviewed simulated CADe prompts of 33 CT colonography studies, 21 of which contained one medium-sized (6–9 mm) adenomatous polyp. Medium-sized polyps are of interest because unaided radiologists tend to miss them more often than larger lesions . Each patient had eight CADe prompts: four in supine and four in prone CT scans. A polyp was indicated either by one true-positive CADe prompt on either CT scan or by a true-positive CADe prompt on both CT scans. According to the results, the use of the matching true-positive CADe prompts yielded a statistically significant 11% increment in the detection sensitivity of computer-assisted readers.
The effect of CADe prompts on observer performance has been studied in medical imaging for more than two decades. A majority of these studies has involved mammography, which is one of the earliest applications of CADe . As with CT colonography, mammography studies have multiple projection images per patient. Retrospective and prospective studies have indicated that the use of CADe prompts can increase breast cancer detection rate , but it has also been observed that CADe systems have low specificity, their sensitivity is relatively poor for subtle lesions, and that CADe prompts do not serve as a decision-making aid . Similar observations have been made with CADe for CT colonography . To date, the single most common complaint of radiologists has been that current CADe systems display too many false positives. Studies and user comments suggest that false-positive CADe prompts increase interpretation time, confuse readers, and that routine observation of a large number of redundant CADe prompts on obviously normal regions can make the reader lose confidence on the ability of the CADe system to be of real help .
It is also well-known that radiologists can dismiss true lesions even if they are indicated correctly by CADe prompts. An often overlooked issue is that even if radiologists can see a lesion, they can still have difficulty in interpreting the image correctly and making the right decision about the case . A recent CADe study in CT colonography indicated that, even though computer-assisted readers reported more of the medium-sized polyps than did unaided readers, the likelihood that a computer-assisted reader incorrectly dismissed a lesion that was indicated correctly by a CADe prompt increased with increasing lesion size and, in particular, with non-polypoid appearance of the lesion .
Obviously, if a reader lacks the training and experience to identify true lesions even if they are pointed out to that reader or if the prompted lesions are not considered “actionable,” CADe prompts alone will be of limited help in improving detection sensitivity. The reader could also dismiss CADe prompts too eagerly because of a conscious or subconscious bias about the CADe system. Such a bias could be caused by a variety of reasons, such as routine display of large numbers of false-positive CADe prompts or expectations that CADe is only useful for indicating certain types of lesions. The reader could also be confused by clinical studies that often generalize their observations made on a specific version of a specific CADe system to apply to all CADe systems, even though another CADe system that uses different detection algorithms could have quite different performance profile with the same or different data .
The conventional approach to improve reader confidence on CADe has been to develop methods for reducing false-positive CADe prompts and for improving detection sensitivity for subtle lesions. However, effective communication between CADe systems and radiologists is becoming increasingly important. To help radiologists make better decisions, CADe systems may need to provide additional information, such as the estimated likelihood that a CADe prompt represents a true lesion; for example, in terms of the size or color of CADe prompts. This would also reduce the gap to computer-assisted diagnosis, in which the goal is to aid radiologists in the diagnosis of already detected lesions but that to date has been kept largely separate from CADe .
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
Get Radiology Tree app to read full this article<
References
1. Chen S.C., Lu D.S., Hecht J.R., et. al.: CT colonography: value of scanning in both the supine and prone positions. Am J Roentgenol 1999; 172: pp. 595-599.
2. Gryspeerdt S.S., Herman M.J., Baekelandt M.A., et. al.: Supine/left decubitus scanning: a valuable alternative to supine/prone scanning in CT colonography. Eur Radiol 2004; 14: pp. 768-777.
3. CT colonography computer-aided polyp detection: effect on radiologist observers of polyp identification by CAD on both the supine and prone scans. Acad Radiol 2010; pp. 948-959.
4. Rockey D.C.: Computed tomographic colonography: current perspectives and future directions. Gastroenterology 2009; 137: pp. 7-10.
5. Giger M.L., Chan H.P., Boone J.: History and status of CAD and quantitative image analysis: the role of medical physics and AAPM. Med Phys 2008; 35: pp. 5799-5820.
6. Birdwell R.L.: The preponderance of evidence supports computer-aided detection for screening mammography. Radiology 2009; 253: pp. 9-16.
7. Taylor P., Champness J., Given-Wilson R., et. al.: Impact of computer-aided detection prompts on the sensitivity and specificity of screening mammography. Health Technol Assess 2005; 9: pp. 1-58.
8. Fisichella V.A., Jäderling F., Horvath S., et. al.: Computer-aided detection (CAD) as a second reader using perspective filet view at CT colonography: effect on performance of inexperienced readers. Clin Radiol 2009; 64: pp. 972-982.
9. Park S.H., Kim S.Y., Lee S.S., et. al.: Sensitivity of CT colonography for nonpolypoid colorectal lesions interpreted by human readers and with computer-aided detection. AJR Am J Roentgenol 2009; 193: pp. 70-78.
10. Summers R.M., Frentz S.M., Liu J., et. al.: Conspicuity of colorectal polyps at CT colonography: visual assessment, CAD performance, and the important role of polyp height. Acad Radiol 2009; 16: pp. 4-14.
11. Taylor S.A., Greenhalgh R., Ilangovan R., et. al.: CT colonography and computer-aided detection: effect of false-positive results on reader specificity and reading efficiency in a low-prevalence screening population. Radiology 2008; 247: pp. 133-140.
12. Petrick N., Haider M., Summers R.M., et. al.: CT colonography with computer-aided detection as a second reader: observer performance study. Radiology 2008; 246: pp. 148-156.
13. Taylor S.A., Charman S.C., Lefere P., et. al.: CT colonography: investigation of the optimum reader paradigm by using computer-aided detection software. Radiology 2008; 246: pp. 463-471.
14. Bielen D., Kiss G.: Computer-aided detection for CT colonography: update 2007. Abdom Imaging 2007; 32: pp. 571-581.
15. Yoshida H., Näppi J.: CAD in CT colonography without and with oral contrast agents: progress and challenges. Comput Med Imaging Graph 2007; 31: pp. 267-284.
16. Philpotts L.E.: Can computer-aided detection be detrimental to mammographic interpretation?. Radiology 2009; 253: pp. 17-22.
17. Roehrig J.: The manufacturer’s perspective. Br J Radiol 2005; 78: pp. S41-S45.
18. Kundel H.L., Nodine C.F., Carmody D.: Visual scanning, pattern recognition and detection. Invest Radiol 1978; 13: pp. 175-181.
19. Kundel H.L., Nodine C.F., Krupinski M.A.: Searching for lung nodules: visual dwell indicates locations of false-positive and false-negative decisions. Invest Radiol 1989; 24: pp. 472-478.
20. Krupinski E.A.: Visual scanning patterns of radiologists searching mammograms. Acad Radiol 1996; 3: pp. 137-144.
21. Mello-Thoms C., Hardesty L., Sumkin J., et. al.: Effects of lesion conspicuity on visual search in mammogram reading. Acad Radiol 2005; 12: pp. 830-840.
22. Taylor S.A., Robinson C., Boone D., et. al.: Polyp characteristics correctly annotated by computer-aided detection software but ignored by reporting radiologists during CT colonography. Radiology 2009; 253: pp. 715-718.
23. Näppi J., Yoshida H.: Adaptive correction of the pseudo-enhancement of CT attenuation for fecal-tagging CT colonography. Med Image Anal 2008; 12: pp. 413-426.
24. Park H.S., Kim S.H., Kim J.H., et. al.: Computer-aided polyp detection on CT colonography: comparison of three systems in a high-risk human population. Eur J Radiol 2010; in press
25. Kim S.J., Moon W.K., Kim S.Y., et. al.: Comparison of two software versions of a commercially available computer-aided detection (CAD) system for detecting breast cancer. Acta Radiol 2010; 51: pp. 482-490.
26. Hein P.A., Krug L.D., Romano V.C., et. al.: Computer-aided detection in computed tomography colonography with full fecal tagging: comparison of standalone performance of 3 automated polyp detection systems. Can Assoc Radiol J 2010; 61: pp. 102-108.
27. Lee M.W., Kim S.H., Park H.S., et. al.: An anthropomorphic phantom study of computer-aided detection performance for polyp detection on CT colonography: a comparison of commercially and academically available systems. AJR Am J Roentgenol 2009; 193: pp. 445-454.
28. Näppi J., Yoshida H.: Automated scheme for preparation-independent detection of colorectal lesions with comparison to conventional CAD in CT colonography. Proc MICCAI 2008 Workshop on Computational and Visualization Challenges in the New Era of Virtual Colonoscopy 2008; pp. 127-134.
29. Gilbert F.J., Astley S.M., Boggis C.R.M., et. al.: Variable size computer-aided detection prompts and mammography film reader decisions. Breast Cancer Res 2008; 10: R72. Available at: http://breast-cancer-research.com/content/10/4/R72 . Accessed June 13, 2010.
30. Näppi J., Frimmel H., Yoshida H.: Virtual endoscopic visualization of the colon by shape-scale signatures. IEEE Trans Inf Technol Biomed 2005; 9: pp. 120-131.
31. Näppi J., Okamura A., Frimmel H., et. al.: Region-based supine-prone correspondence for the reduction of false-positive CAD polyp candidates in CT colonography. Acad Radiol 2005; 12: pp. 695-707.
32. Suh J.W., Wyatt C.L.: Deformable registration of supine and prone colons for computed tomographic colonography. J Comput Assist Tomogr 2009; 33: pp. 902-911.
33. Paquerault S., Petrick N., Chan H.P., et. al.: Improvement of computerized mass detection on mammograms: fusion of two-view information. Med Phys 2002; 29: pp. 238-247.
34. van Engeland S., Karssemeijer N.: Combining two mammographic projections in a computer aided mass detection method. Med Phys 2007; 34: pp. 898-905.
35. Huang A., Roy D.A., Summers R.M., et. al.: Teniae coli-based circumferential localization system for CT colonography: feasibility study. Radiology 2007; 243: pp. 551-560.
36. Wei J., Chan H.P., Sahiner B., et. al.: Computer-aided detection of breast masses on mammograms: dual system approach with two-view analysis. Med Phys 2009; 36: pp. 4451-4460.
37. Velikova M., Samulski M., Lucas P.J., et. al.: Improved mammographic CAD performance using multi-view information: a Bayesian network framework. Phys Med Biol 2009; 54: pp. 1131-1147.
38. Burling D., Moore A., Marshall M., et. al.: Virtual colonoscopy: effect of computer-assisted detection (CAD) on radiographer performance. Clin Radiol 2008; 63: pp. 549-556.
39. Nishikawa R.M., Edwards A., Schmidt R.A., et. al.: Can radiologists recognize that a computer has identified cancers that they have overlooked?. SPIE Medical Imaging 2006; 614601:1–8