Mammographie image segmentation using kernel based FCM clustering approach

Breast cancer is the second leading causes of cancer related deaths among women. Detection of breast cancer at the initial stages increases the probability of survival of the patient. Mammography has been one of the most reliable methods for early detection of this disease. Computer-aided diagnosis (CAD) of mammograms has received great attention because of its speed, consistency and providing a better solution for automatic detection of breast cancer. However, poor visibility of mammographic image addresses the necessity of accurate segmentation technique. In this paper we have introduced a novel and advanced kernel based fuzzy c-mean (FCM) clustering technique for segmentation of mammographic masses. To improve the accuracy of the segmentation process two important parameters that convey the properties of masses like ‘entropy’ and “intensity mean” of the kernel are taken into account for fuzzification purposes. This kernel is moved across the mammograms to collect all possible values and hence these feature values are exploited as the data of FCM clustering technique. We have also compared the performance of the proposed approach with intensity based conventional FCM clustering technique and have found better segmentation results for both visual interpretation as well as high level detection.