Filtering False Detections of Small Multiple Sclerosis Lesions using Fuzzy Regional Analysis

Abstract

This paper introduces a method to filter false detections of small multiple sclerosis lesions in magnetic resonance images based on the analysis of regional features. The proposed method considers as starting point the results of an earlier work in which, through the use of fuzzy rules, the image pixels showing hyperintensity were detected. The regional analysis of the results obtained at previous work allows extracting some features with differentiation capability between small multiple sclerosis lesions and false detections. These features are introduced as restrictions for obtaining a new and improved fuzzy membership function associated with the presence of hyperintensity in these images. Results show an important reduction of the number of false detections preserving the small multiple sclerosis lesions previously detected.