Improvement of Pseudo Multispectral Classification of Brain MR Images

Classification of brain tissues is becoming an increasingly useful tool for investigating the aging brain or disease-induced brain alterations. Numerous strategies have been proposed to classify brain tissues into white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF). However, many of them fail when classifying specific regions with low contrast between tissues (e.g. cerebellum gray and white matter). Erroneous classification may lead to volume overor under-estimation, thus leading to equivocal interpretations. In this work, instead of using gray scale images (T1-MPRAGE) to classify brain tissues, we use an improved pseudo multispectral classification (PMC) technique using CIE XYZ spaces and iterative K-mean clustering in order to enhance classification of the brain GM, WM and CSF. The accuracy of the proposed approach is assessed using atlas brain templates and compared with FSL classification (FMRIB, Oxford, UK). Keywords— Classification, K-mean, Clustering, CIE XYZ.