Jointly using computationally selected and clinically suggested cortical volumes for automated identification of mild cognitive impairment

Mild cognitive impairment (MCI) has been widely seen as the prophase of Alzheimer's disease, the most prominent kind of dementia, which has become a global health problem and social threat due to its damage to the cognitive function. Magnetic resonance imaging (MRI) offers the ability to visualize degenerative histological changes, and hence has been widely used to diagnose MCI from normal aging. In this paper, we use statistics to characterize each cortical volume obtained by spatially normalizing the brain MRI study onto the automated anatomical labelling (AAL) cortical parcellation map, and adopt the integer-coded genetic algorithm (GA) to computationally select cortical volumes, based on which accurate diagnosis of MCI can be achieved. Our results suggest that the 17 cortical volumes recommended by medical professionals underperform the 17 volumes selected by GA and jointly using the volumes, which were recommended simultaneously by clinicians and GA, and those, which were selected repeatedly by GA in different settings, can further improve the accuracy of MCI differentiation.