AI model predicts ALS progression and wheelchair need with digital twin approach

Researchers from Temple University and Drexel University have developed a temporal machine learning framework that predicts clinically meaningful milestones in ALS (amyotrophic lateral sclerosis) patients. The model, described in a preprint on arXiv, uses a digital-twin-inspired approach combining longitudinal ALSFRS-R (ALS Functional Rating Scale-Revised) trajectories with survival modeling to forecast functional decline and assistive device utilization.

The team constructed a harmonized longitudinal dataset integrating diagnosis records, ALSFRS-R assessments, activities of daily living, and demographics. They applied correlation-based clustering to identify five coherent functional domains — bulbar, upper limb, axial, lower limb, and respiratory. Generalized additive mixed models characterized nonlinear decline across these domains. Using Cox proportional hazards modeling, they found that lower limb function, particularly walking and stair climbing, was the strongest predictor of earlier wheelchair access.

Building on these insights, the researchers implemented a digital twin-inspired temporal time-to-event (TTE) model that generates individualized survival curves and dynamically predicts wheelchair-free survival. The framework is designed to be scalable, interpretable, and clinically actionable — enabling personalized decision support for proactive care planning, clinical trial stratification, and precision medicine in ALS. This work represents a step toward integrating machine learning with neurodegenerative disease management.