In this paper, we describe an improvement of 3D-IMPRESS-Pro which can construct a three dimensional image processing procedure to detect abnormal regions considering the requirement on sensitivity and the number of false positive regions. The modules for feature measurement and classification of suspicious regions were modified so as to be able to construct the procedure which meets higher requirement. We applied the improved system to the problem of generating a lung cancer detection procedure. The paper presents results of applying the procedure to real three dimensional chest CT images and evaluating its performance.

View full abstract