A new feature extraction method for identification of affected regions and diagnosis of cognitive disorders

Cognitive disorders like AD progressively disintegrate neurons and their interconnections in the brain; thus gradually deteriorating cognitive functions. Automated diagnosis is very important in the early diagnosis of cognitive disorders. Early diagnosis allows in taking measures helping the person to move on. Clinical diagnosis is inefficient as the symptoms start to manifest only after significant atrophy of the cortical structures. This makes management of the conditions difficult. Resent findings have revealed the potential of Neuroimaging as a highly effective tool in the early detection of these disorders as structural changes in the brain set in much before the manifestation of observable symptoms. The disorders are a reflection of degeneration of the cortical structures and hence can be detected by analysis of the structural images of the brain. Therefore, analysis of T1-weighted MRI has become a popular method of early diagnosis of AD. The work proposes a feature extraction method that enables simultaneous identification of the afflicted cortical structures and diagnosis of disorders. The method proposed is based on sparse logistic regression and linear discriminant analysis. The results obtained were better than or comparable with many of the works reported in literature.