False-positive reduction in computer-aided detection for bladder tumor using 2D flattening method and shape analysis

High false-positive (FP) rate is always one of the major problems to be solved in Computer-aided detection (CAD) scheme for bladder cancer. In this paper, we propose a novel method for CAD scheme to reduce FP. We atten the three-dimensional (3D) virtual bladder model with suspicious regions highlighted constructed in CAD scheme to a two-dimensional (2D) image. On the attened image, two parameters reecting the shape distinctions between FP and True Positive (TP) are calculated for each suspect region, based on which FPs are eliminated. This method was tested on patients' dataset with different shapes of FPs and received favorable results.