Researchers have developed a new multi-task learning framework called Orthogonal Task Decomposition (OrthTD) to better disentangle shared and task-specific signals from multimodal clinical data. This approach uses a unified Transformer for fusion and imposes an orthogonality constraint to reduce redundancy and isolate specific patient representations. Evaluated on a cohort of 12,430 surgical patients for predicting four outcomes, OrthTD achieved an average AUC of 87.5% and an average AUPRC of 37.2%, outperforming existing methods, particularly in identifying rare events. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT This method could improve the accuracy of predicting multiple outcomes from complex clinical datasets, especially for rare events.
RANK_REASON The cluster contains an academic paper detailing a new method for multi-task learning in clinical data.