Adaptive Fuzzy C-Means Algorithm using the Hybrid Spatial Information for Medical Image Segmentation

Abstract

This paper presents a technique for incorporating different forms of spatial information into the conventional FCM. New modified version of the standard FCM function and a weighted one has been added together to from the modified objective function.. The Euclidian distances are improved to account for the distances of the neighboring pixels. In this hybrid algorithm, the addition of the local spatial information and the modification of the membership are applied in separate steps. However, the distances are computed by replacing the pixel by its neighborhood average to reduce additive noise. Results of clustering and segmentation of synthetic and simulated medical images are presented to compare the performance of the new modified algorithm of hybrid spatial information (HFCM) with the conventional FCM, local spatial information based FCM (SFCM), local membership based FCM (LMFCM), and the Robust spatial data based FCM (RFCM)