In this paper, we report the image processing technique of the computer-aided diagnosis system for the Lung Cancer Screening CT (LSCT). This new LSCT system has many image data, about 30 slices per person. To solve this difficult problem, we tried to reduce the image information drastically to be displayed for the doctor, by image processing techniques. We investigated the system that used the following two methods in parallel, and a satisfactory result was obtained. (1) We developed a modified Maximum Intensity Projection method. By this method, three-dimensional information composed of about 30 slices per person was projected to two-dimensional plane without loss of information of the nodules. (2) We developed anautomatic recognition method of candidates for the nodules, named the Variable N-Quoit filter. By this method, the doctor will only check the CT images where a candidate shadow was automatically picked up.

View full abstract