Minkowski functionals based brain to ventricle index for analysis of AD progression in MR images

In this work, Minkowski functionals (MFs) based brain to ventricle index is proposed to quantify the structural changes in Alzheimer’s MR brain images for severity detection. Initially, the original images (N = 120) from MIRIAD and ADNI database, are skull stripped using morphology based method. The ventricles are segmented using localized region-based active contour method. These algorithms are validated using performance measures. The whole brain and the segmented ventricles are subjected to analysis using MFs. The Minkowski Formulation in 2D (MF2D) of whole brain, ventricle and ratio of brain to ventricle index are formulated. The prominent index is correlated with the Mini-Mental State Examination (MMSE) score. Results show that the correlation of the extracted whole brain (R = 0.80) and the ventricle (R = 0.82) with the experts ground truth is high. It is observed that the MF2D brain to ventricle index provide better discrimination of normal and abnormal subjects (p < 0.0001). Its correlation with MMSE is observed to be high for normal and very high for Alzheimer subjects. Hence this index captures the structural changes and could be used to quantify the progression in neurodegenerative disorder such as AD.