A multi‐modal, multi‐atlas‐based approach for Alzheimer detection via machine learning

• DOI: 10.1002/ima.22263
• Publication Date: 2018-01-10T17:24:39Z
• Authors: Yousra Asim; Basit Raza; Ahmad Kamran Malik; Saima Rathore; Lal Hussain; Mohammad Aksam If...
• Keywords: Brain atlas; Spatial normalization

Abstract The use of biomarkers for early detection Alzheimer's disease (AD) improves the accuracy imaging‐based prediction AD and its prodromal stage that is mild cognitive impairment (MCI). Brain parcellation‐based computer‐aided methods detecting MCI segregate brain in different anatomical regions their features to predict MCI. parcellation generally carried out based on existing atlas templates, which vary boundaries number regions. This works considers dividing atlases combining extracted from these parcellations a more holistic robust representation. We collected data ADNI database divided brains two well‐known atlases: LONI Probabilistic Atlas (LPBA40) Automated Anatomical Labeling (AAL). used baselines images structural magnetic resonance imaging (MRI) 18 F‐fluorodeoxyglucose positron emission tomography (FDG‐PET) calculate average gray‐matter density relative cerebral metabolic rate glucose each region. Later, we classified AD, cognitively normal (CN) subjects using individual template combined both atlases. reduced dimensionality principal component analysis, support vector machines classification. also ranked mostly involved classification determine importance accurately classifying subjects. Results demonstrated calculated multiple lead improved performance compared those one only.