Feature Selection Optimization Using Artificial Immune System Algorithm for Identifying Dementia in MRI Images

Dementia is a common neurodegenerative disease. Magnetic Resonance Imaging (MRI) is widely used for diagnosing dementia. Classification to diagnose neuroimaging issues are automated as standard clinical decisions are quicker, and unaffected by individual neuro-radiological opinions. Automatic dementia classification of MRI medical images using machine learning techniques is presented in this paper. For evaluation, MRI images from OASIS dataset are used. MRI images are segmented and features are extracted from segmented image using Discrete Wavelet Transform. Feature selection is via proposed Artificial Immune System (AIS), that searches solution space for correlation based feature selection. Naive Bayes, CART, C4.5 and K nearest neighbour then classifies the selected features as dementia or non-dementia.