Computer-aided cancer classification system using a hybrid level-set image segmentation

Abstract

In computer vision, image segmentation plays an indispensable role. Image segmentation is used to find objects and their boundary limits in images. In medical imaging, tasks such as finding pathological regions and automatic detection of diseases are a complicated problem for computer science and image processing. This problem is intricate because of insufficient clarity of images. High segmentation accuracy, efficiency, reliability, and ability to handle noise or inhomogeneous intensity are some of the desired characteristics of a good segmentation algorithm for medical images. This study proposes an automatic computer-aided system that has been developed by using a level-set region-based active contour segmentation algorithm for edge detection of 2D medical images with and without inhomogeneous intensity. In this method, cancer images are segmented in the transformed domain, which helps to reduce the noise. The proposed segmentation algorithm combines the local geometric properties and local image fitting properties to correctly identify the pixels corresponding to different regions, in images with or without intensity inhomogeneity. The performance of the proposed method is analyzed with panoramic radiograph, Matrix Laboratory. In the experiments, the proposed method outperforms the baselines.