Despite the importance of the detection of cancers, only a very small number of CAD schemes have been developed for detection of colorectal masses that are likely to be cancers. This is probably because colorectal masses are generally considered to be easily seen by radiologists due to their size and invasiveness. On the contrary, it is not easy for CAD to detect and accurately delineate entire mass regions; instead, CAD tends erroneously to report local surface bumps of a mass as several polyps.
The detection of both polyps and masses by CAD would be a more efficient computer aid in the interpretation of CTC examinations than is the detection of polyps alone. If masses are not detected by CAD, radiologists need to perform a careful and complete review of all CTC cases for the presence of masses, which may increase the reading time. Moreover, accurate detection of masses may depend on radiologists' experience and on how rapidly they read the cases (Morrin et al. 2003). Therefore, the application of CAD to the detection of masses could improve the diagnostic accuracy of CTC by reducing potential reading errors due to reader fatigue, inexperience, or a too rapid reading. Furthermore, without explicit mass detection, CAD could also confuse radiologists by presenting portions of masses as several polyps.
Automated detection ofmasses poses challenges for CAD because they may appear as intraluminal types (lobulated, polypoid, or circumferential) or nonin-traluminal types (mucosal wall-thickening type of growth pattern or masses that block the colon), both of which have a wide variation in shape characteristics. Only a few CAD schemes for the detection of colorectal cancers have addressed this challenge (Nappi et al. 2002b, 2004a). One of these studies (Nappi et al. 2004a) used a fuzzy merging method and wall-thickening analysis for delineation of intraluminal and non-intraluminal masses, respectively. The CAD scheme detected 93% of masses (13 of the 14 masses) in 82 patients and extracted their regions, with 0.21 false positives per patient on average. Figure 11.8a shows a 50-mm intraluminal circumferential mass with apple-core morphology, and Figure 11.8b shows its endoscopic view. The entire mass region was extracted by the mass detection method, as indicated by the white regions in Figure 11.8c,d.
Preliminary results indicate that CAD has the potential to detect colorectal masses in CTC with high accuracy. However, further research and a large-scale evaluation are needed for development of a CAD scheme that detect and delineate various types of masses reliably.
Was this article helpful?
Stop The Pain. Manage Your Diverticular Disease And Live A Pain Free Life. No Pain, No Fear, Full Control Normal Life Again. Diverticular Disease can stop you from doing all the things you love. Seeing friends, playing with the kids... even trying to watch your favorite television shows.