Fast and accurate pseudo multispectral technique for whole-brain MRI tissue classification

Abstract

Brain imaging using high resolution magnetic resonance imaging (MRI) has expanded the role of neuroimaging in clinical and research setting, particularly when used in junction with advanced data extraction and visualization tools. The accurate interpretation of information pertaining to investigate the aging brain or disease induced brain alterations, as well as functional studies of brain tissues relies on accurate and reproducible brain tissue classification, which is the preliminary step prior to segmentation (Pham et al 2000, Sonka and Fitzpatrick 2000). Manual tissues classification may be reliable, but is difficult to perform in clinical routine and is time consuming (Hatt et al 2017). Furthermore, it suffers from large intra-and inter-observer variability that decreases the credibility of segmentation-based image analysis (Kasiri et al 2010). As such, there is a large demand to develop alternative methods for robust and accurate automated classification of brain MR images as a prerequisite for