Hybrid Detection Model Combining Grey Scale Weight Based Quoit Filter with Support Vector Machine

This paper is based on our previous study of automatic liver segmentation, and proposes an automatic approach for the detection of near-spherical liver tumors. Firstly, an improved variable quoit filter model is presented, which utilizes the characteristics of the ring-shaped filter to detect the rotationally symmetric structure. Secondly, in order to improve the detection performance of quoit filter, the sensitivity of multi-size tumor is enhanced by a specially designed grey-scale weight-based distance transformation function. Finally, support vector machine classification algorithm is adopted to reduce the false positive of classification. Experiment on 20 sets of abdominal CT images shows that the modified filter can achieve a better performance than the conventional filter, which indicates a promising prospect for spherical liver tumors detection.