Automated Machine Learning Approach (AutoML) to Alzheimer’s Disease Diagnosis and Prognosis

Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by memory loss. While applying Machine Learning (ML) demands a certain level of expertise, which is often a barrier for healthcare professionals, automated machine learning (AutoML) significantly lowers this barrier. This study analyzes an AutoML tool (PyCaret) for AD classification and prediction. Two experiments were designed to evaluate its diagnostic and prognostic capabilities with AD, Mild cognitive impairment (MCI), and Normal Controls (NC). SHapley Additive exPlanations (SHAP) was used to explain the ML models. For diagnosis, it had an accuracy of 98.6% for NC vs AD, 91.3%, for NC vs MCI, 92.5% for MCI vs AD, and 89.5% for the multiclass NC vs MCI vs AD. Regarding the prognosis capabilities, prediction of future cognitive states four years after their initial visit produced an accuracy of 92.8% for NC vs AD, 82.7% for NC vs MCI, 90.2% for MCI vs AD, and 81.4% for NC vs MCI vs AD. These results are in range and, in some cases, improve the state of the art even when compared to deep learning solutions. They confirm the potential of AutoML tools to automate ML algorithm selection and tuning for a specific medical application.