Classification of Brain Neuroimaging for Alzheimer's Disease Employing Principal Component Analysis

— Alzheimer's disease (AD) is one illness that significantly impacts people’s lives. As AD worsens over time, it causes the death of brain cells. To assist a neurologist, a proposed classification method for AD progression is introduced in this paper. Pre-processing is applied to clean up artifacts from brain images. As biomarkers for AD diagnosis, three specific areas of the brain are utilized. Multiplicative intrinsic component optimization with an exemplar pyramid is employed for the three main biomarkers segmentation at a multi-scale. For feature extraction, the gray-level co-occurrence matrix is utilized. Finally, principal component analysis is incorporated for feature reduction, and based on the Euclidean distance the decision of the binary classifier is performed. The Alzheimer's Disease Neuroimaging Initiative baseline dataset is used with 311 subjects, 262 for training and 49 for testing. The proposed method achieved an accuracy of 96.296% for the classification between late mild cognitive impairment (LMCI) and cognitive normal (CN), 85.71% between early mild cognitive impairment (EMCI) and CN, 92% between AD and CN, 95.833% between EMCI and LMCI, 91.3% between AD and EMCI, and 84.21% between AD and LMCI. Evaluation results show that the proposed method enhanced the existing method's accuracy with less feature dimensionality.