Masahiro Yakami MD, PhD, Presenter: Nothing to Disclose

Masami Kawagishi, Abstract Co-Author: Nothing to Disclose

Gakuto Aoyama, Abstract Co-Author: Nothing to Disclose

Koji Fujimoto MD, PhD, Abstract Co-Author: Nothing to Disclose

Takeshi Kubo MD, Abstract Co-Author: Nothing to Disclose

Kaori Togashi MD, PhD, Abstract Co-Author: Research Grant, Bayer AG Research Grant, DAIICHI SANKYO Group Research Grant, Eisai Co, Ltd Research Grant, FUJIFILM Holdings Corporation Research Grant, Nihon Medi-Physics Co, Ltd Research Grant, Shimadzu Corporation Research Grant, Toshiba Corporation Research Grant, Covidien AG

Ryo Sakamoto, Abstract Co-Author: Nothing to Disclose

Koji Sakai, Abstract Co-Author: Nothing to Disclose

Hiroyuki Sekiguchi, Abstract Co-Author: Nothing to Disclose

Yutaka Emoto MD, PhD, Abstract Co-Author: Nothing to Disclose

Yoshio Iizuka, Abstract Co-Author: Nothing to Disclose

Hiroyuki Yamamoto, Abstract Co-Author: Nothing to Disclose

Many CADx systems have been reported to improve differential diagnosis on lung nodules by radiologists. However, radiologists still have reluctance to accept CADx suggestion. To improve diagnostic accuracy with a CADx system, it is also important to improve persuasiveness of CADx suggestion, as well as the diagnostic accuracy. Thus we developed a CADx system which suggests the diagnosis on a specified lesion and reasons for the diagnosis, and evaluated the persuasiveness of the suggestion.

With the approval of the institutional review board, we built a database on 491 lung nodules on which diagnoses were clinically confirmed as primary lung cancer, metastatic nodules or other benign nodules. This database consisted of CT images, image findings on the nodules, the confirmed diagnosis, clinical information such as laboratory data and patient history. The image findings were described by consensus of two board-certified radiologists. The CADx was trained and evaluated by using 179 and 312 nodules in the database, respectively. The CADx derived and suggested a list of possibilities for differential diagnoses on each nodule using a Bayesian network (ICAD). It also derived image findings and/or clinical information having high influence on the diagnosis with the highest possibility and suggested them as the reasons for the inference in addition to the list (RCAD). Eleven radiologists, with five years’ experience in diagnostic imaging, interpreted the 312 nodules under three different conditions (without CAD, with ICAD, with RCAD) with more than one month intervals. The numbers of cases on which each radiologist disagreed with the CADx initially, and changed his/her diagnosis to follow the CADx suggestion, were counted as disagreed and “persuaded” cases for evaluation, respectively.

The average number of disagreed cases among the 11 radiologists were 99.5 (SD=13.5). That of “persuaded” cases by RCAD among them was 47.5 (SD=15.2), and significantly larger than that by ICAD, 43.9 (SD=13.9) (Wilcoxon signed-rank test, p<0.05).

RCAD was more persuasive for the radiologists than ICAD.

Yakami, M, Kawagishi, M, Aoyama, G, Fujimoto, K, Kubo, T, Togashi, K, Sakamoto, R, Sakai, K, Sekiguchi, H, Emoto, Y, Iizuka, Y, Yamamoto, H, Improving Persuasiveness of Computer-aided Differential Diagnosis (CADx) System by Disclosing Reasons for Diagnosis.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13016987.html