For patients with mild cognitive impairment (MCI), an early stage of Alzheimer’s Disease (AD), monitoring the risk of progressing to AD over time is crucial. Recent research published in JAMIA Open has introduced a state-of-the-art deep learning framework named HiMAL (Hierarchical Multi-task Auxiliary Learning) that addresses this need.
The study, led by Sayantan Kumar, PhD Candidate, Thomas Kannampallil, PhD, and Philip Payne, PhD, along with Sean Yu, PhD, and Andrew Michelson, MD, outlines how HiMAL can predict whether an MCI patient will progress to AD within the next six months. This prediction is made at multiple points along the disease trajectory, providing a timely and dynamic assessment.
HiMAL’s capabilities extend beyond prediction. It offers clinically informative model explanations that anticipate cognitive decline six months in advance. These insights are invaluable for clinicians assessing future disease progression and planning appropriate interventions.
The framework is built on routinely collected Electronic Health Records (EHR) data, highlighting its practical application for point-of-care monitoring and management of high-risk AD patients. HiMAL’s translational potential could significantly impact how Alzheimer’s Disease is managed in clinical settings.