Diagnosis of breast cancer by optical image analysis

We consider the process of object detection, recognition and classification in digital optical images of human breast cells with the aim of differentiating between normal and abnormal (cancerous) cells. The work is based on research into the development of a breast cancer screening system that can be used by cytologists to differentiate between benign and malignant types using images that are typical of those currently interpreted by cytologists world-wide. The approach considered is based on feature vectors which are of two types. We consider statistical features such as the mode, median, mean, and standard deviation and features composed of Euclidian geometric parameters such as the object perimeter, area and infill coefficient. All components of the feature vectors are computed to `reflect' the statistical characteristics and the geometric structure of the imaged cells. The recognition process includes a segmentation algorithm based on an adaptive imaging threshold procedure that is sensitive to local ranges in pixel intensity (minimum-maximum values). Decision criteria are based on the application of Fuzzy Logic and Membership Function theory. In particular, we present a technique for the creation and extraction of data to construct the Membership Function.