VLSM-Net: A Fusion Architecture for CT Image Segmentation

Region of interest (ROI) segmentation is a key step in computer-aided diagnosis (CAD). With the problems blurred tissue edges and imprecise boundaries ROI medical images, it hard to extract satisfactory ROIs from images. In order overcome shortcomings V-Net model or level set method (LSM), we propose this paper new image method, VLSM-Net model, combining these two methods. Specifically, first use segment ROIs, result as initial contour. It then fed through hybrid LSM for further fine segmentation. That is, complete can be obtained by successively LSM. The experimental results conducted public datasets LiTS LUNA show that, compared with alone, our greatly improves sensitivity, precision dice coefficient values (DCV) 3D segmentation, thus validating model’s effectiveness.