Lesion segmentation in acute cerebral infarction based on Dempster-Shafer theory

Abstract

In the diagnosis and treatment of acute cerebral infarction, it will be helpful for doctors to implement disease assessment and develop treatment plans if infarct and cytotoxic brain edema around the infarct can be observed and distinguished. In this paper, a method of fuzzy c-means clustering combined with Dempster-Shafter theory is used to achieve lesion segmentation by combining information from two different modalities of magnetic resonance imaging. The basic probability assignment function of each image type is obtained from membership degrees of all image pixels in image using fuzzy c-means clustering method. Dempster-Shafer combination rule is then applied on different basic probability functions corresponding to the modal images to decrease uncertainty and conflicting information. The results show that infarct and cytotoxic brain edema around the infarct can be distinguished in the final segmentation map, and that the size and outline of the edema area are accurate, which will help doctors diagnose and assess situation of patients with acute cerebral infarction.