Application of Active Region Model for Detection of Liver Cancer

The most difficult procedure of image processing [3] is the task of image segmentation. The popular but well-motivated opinion is that medical images are particularly hard to analyze [1]. One of the most promising methods of image segmentation is the active contour approach [2,4,5,7-11]. Originally, active contour methods were developed as tools for low-level image segmentation but with the ability to use high-level information. The high-level information is hidden in the given objective function, called energy, which is used for evaluation of the quality of a contour generated by the method. As shown in [8,10], contours are contextual classifiers of pixels (one part of the pixels belongs to the interior and the other part to the exterior of a given contour) and active contours are methods of optimal construction of classifiers. The concept of active region model is similar [4]. The difference lies in the evaluation of the quality of the segmentation result. While in the active contour methods, it is usually the shape of the contour that is examined, the active region models additionally evaluate the statistical parameters of the area inside the examined contour. The paper is composed in the following way. First, the applied segmentation method, called active regions, is briefly presented. Then, the method is examined using liver cancer images. Because of its features, the method can be applied to automatic scanning of liver images. It must be emphasized that further the active contour term will be used as a synonym for snakes [5] method although the presented approach can be applied to other variants of active contour method.