Multimodal diagnosis of Alzheimer's disease based on volumetric and cognitive assessments

Rashmi Kumari, Subhranil Das, Akriti Nigam, Raghwendra Kishore Singh

Abstract: Alzheimer’s Disease (AD) is the most prevalent form of dementia, severely impacting memory functions. Early classification of AD patients, distinguishing them from those with Mild Cognitive Impairment (MCI) or normal cognition (NC), is crucial for clinical intervention. However, traditional methods relying on single MRI feature evaluation yield suboptimal accuracy. Recent advances in multimodal studies show that combining multiple features from various T1-weighted MRI (T1-W1) images can significantly enhance classification accuracy. In this study, we propose a novel Decision Classification Tree (DCT) technique applied to T1-weighted MRI images, incorporating cognitive assessment for global volume measurement. It utilizes morphometric features and spatial information from different Regions of Interest (ROI) within localized brain areas, improving classification accuracy. We evaluate DCT against seven state-of-the- art techniques for binary classification categories (AD vs MCI, MCI vs CN, and AD vs CN) using metrics like accuracy, sensitivity, specificity, and ROC curve analysis. Testing on the ADNI dataset with 102 subjects, we achieve an accuracy of 91.6%, 100% specificity, and 83% sensitivity for the AD vs CN binary class. For MCI vs AD, we attain 76% accuracy, 77.2% specificity, and 80% sensitivity. For MCI vs CN, our model achieves 80% accuracy, 89.4% specificity, and 80% sensitivity.

External IDs:dblp:journals/mta/KumariDNS25