Aim: Dementia is a common neurodegenerative disease which propagates itself through minute symptoms and develops into a form of severe brain damage. Magnetic Resonance Imaging (MRIs) are currently the best medical imaging tools that permit cross-sectional views of the human body with excellent tissue contrasts. MRIs play a significant part in the appraisal of pathological features of brains and are effective in diagnosis of dementia. In this work, the brain MRI are classified as dementia and non-dementia. Features are extracted using Discrete Wavelet Transform (DWT) and feature selection is via proposed hybrid Artificial Immune System (AIS). Genetic Algorithms (GAs) are combined with AIS to optimize the feature subset selection. Naïve Bayes, C4.5 and K nearest neighbour then classifies the selected features as dementia or non-dementia. Experimental results show that the proposed method is effective in improving the efficiency of the classifiers. © 2017 Elsevier B.V., All rights reserved.