MRI-Based Hippocampal Radiomics for Early Diagnosis and Staging of Alzheimer's Disease

Motivation: Alzheimer's disease (AD) poses a significant challenge due to global aging, characterized by complex neurodegenerative processes. Goal(s): To investigate MRI-based hippocampal radiomics as diagnostic markers for AD and differentiate mild cognitive impairment (MCI) from dementia in AD. Approach: We integrate T1-weighted imaging (T1WI) with machine learning to develop AD diagnostic and differentiation models. Results: The Random Forest model exhibited the highest classification diagnostic performance, with the highest areas under the curve (AUC) of 0.977 for AD vs. HC and 0.908 for AD-MCI vs. AD-D. Decision curves analysis indicated good clinical utility of the models at probability thresholds between 0.2 and 0.8. Impact: The potential of T1WI-based hippocampal radiomics features for diagnosing AD is emphasized in this study. Our findings provide enhanced insights into AD diagnosis, progression monitoring, and potential management guidelines.