Detection of hyperintense regions on MR brain images using a Mamdani type Fuzzy Rule-Based System: Application to the detection of small multiple sclerosis lesions

In this paper we present an algorithm for detecting hyperintense regions in brain images acquired by Magnetic Resonance Imaging. The work is part of a more general research oriented to the design of support tools that assist the healthcare experts in their research activities on brain diseases. The algorithm has been focused on the detection of small multiple sclerosis lesions in PD- and T2-weighted images. In the design of the algorithm we have considered a fuzzy approach to deal with the uncertainty and vagueness characteristic of these lesions in magnetic resonance images. The core of the work is the introduction of a Mamdani type Fuzzy Rule-Based System to optimize the detection taking into account the necessary trade-off between true and false positives in this kind of problems. Results show a very good sensitivity of the algorithm in the detection of hyperintense regions associated with small multiple sclerosis lesions, and a low false positive rate with regard the number of pixels analyzed.