Mitsuru Sato, Presenter: Nothing to Disclose

Toshihiro Ogura PhD, Abstract Co-Author: Nothing to Disclose

Norio Hayashi PhD, Abstract Co-Author: Nothing to Disclose

Yoshiaki Yasumoto, Abstract Co-Author: Nothing to Disclose

Hyunjong Lim, Abstract Co-Author: Nothing to Disclose

Kunio Doi, Abstract Co-Author: Nothing to Disclose

Our computerized method based on a template-matching technique would be useful in assisting radiologist in the detection of lymph nodules adjacent to colon cancer in CTC.

A major challenge in the current computer-aided detection (CAD) in CT colonography (CTC) is to detect polyps at a high sensitivity level. However, there is almost no reports regarding the detection of swelling lymph nodes on CTC images. Our purpose was to develop a novel CAD method for detection of swelling lymph nodes adjacent to tumor using a template-matching technique on CTC images.

For detection of swelling lymph nodes, our method was applied to reconstructed slab maximum-intensity-projection (MIP) images in orthogonal plane of virtual endoscopic images. Since CT artifacts of MIP images can be reduced compared with that of MPR images, CAD system can be developed for detection of round patterns on relatively low noise images. Our approach consisted of the following steps: 1) creation of templates by use of swelling lymph nodes, 2) initial candidate detection using template matching technique and, 3) reduction of false positives using image features. Our scheme was evaluated with use of FROC analysis.

Since MIP images provided extra-luminal images by depicting blood vessels as liner patterns, blood vessels were removed easily. Since this method was applied to a large number of MIP images, it is possible to have a very high sensitivity in the detection of swelling lymph nodes. The initial candidate detection before classification achieved a 100% (14/14) sensitivity in the detection of peripheral lymph nodes with 12 FPs per case. The subsequent feature selection removed 66.7% (8/12) of FPs without removal of lymph nodes in a leave-one-lesion-out cross-validation test; thus, a 100% sensitivity with 4 FPs per patient was achieved.

Sato, M, Ogura, T, Hayashi, N, Yasumoto, Y, Lim, H, Doi, K, Computerized Method for Detection of Swelling Lymph Nodes Adjacent to Colon Cancer on CTC Images: Preliminary Results.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13012587.html