Analysis of Anatomical Regions in Alzheimer's Brain MR Images Using Level Sets and Minkowski Functionals

Abstract

In this work, an attempt is made to analyze anatomical regions of Alzheimer's brain MR images using level set and Minkowski functionals (MFs). The T1 weighted sagittal view of normal and abnormal images considered in this work, are obtained from MIRIAD database. The ventricle along with the hippocampus is segmented using Reaction Diffusion (RD) level set method and is subjected to further analysis using MFs. The prominent feature derived from the segmented region which contains ventricle and hippocampus is correlated with the MiniMental State Examination (MMSE) score. Results show that RD method is able to delineate the exact boundary of the ventricle along with hippocampus. It is observed that the MF-area of segmented region provides better discrimination of normal and abnormal subjects (p < 0:001) compared to the MF-perimeter and MF-Euler number. The correlation of the MMSE with prominent MF-area for normal, mild, moderate and severe are found to be 0.71, 0.82, 0.84 and 0.89, respectively. This suggests that characterization of brain structures using MFs improve the discrimination capability of brain disorders. Hence MF-area feature could be used for the study of progression in Alzheimer's disease (AD) like neurodegenerative disorders.