Abstract
PurposeCombined with Diffusion-Weighted imaging, functional MRI, and electroencephalography, hemisphere's asymmetries made it possible to understand morphological brain changes due to Alzheimer's disease. However, it is not used sufficiently in association with Structural MRI. In this article, we evaluate the efficiency of the score's asymmetry of different regions of interest in combination with machine learning algorithms for the diagnosis of Alzheimer's disease in his earlier stage.MethodsThis study examines a 275 T1-weighted brain: 82 normal controls (NC), 52 Early Mild Cognitive Impairment (EMCI), 70 Late Mild Cognitive Impairment (LMCI), and 71 AD patients. A framework has been performed to visualize the accuracy of five classifier algorithms in response to different selected features. This procedure has been performed with voxel-based morphometry (VBM) of regions of interest (ROI) and asymmetry scores. Due to his highest performances, random forest has been selected to establish and evaluate the multiclassification separately with the two types of features. Finally, results have been compared and anatomical regions affiliated to relevant asymmetry scores have been analyzed.ResultsEven if VBM features were more efficient in the classification of MCI and AD among CN, although for the discrimination between LMCI and EMCI, all the evaluation metrics based on asymmetry scores are the highest for the differentiation between CN and EMCI cohorts.ConclusionOverall, using the asymmetry scores has proved efficient in the discrimination of the EMCI cohort. Although, Amygdala asymmetry has been identified as a biomarker of disease at the earlier stage.