Automated Fusion of Multimodal Electronic Health Records for Better Medical Predictions
This addresses the problem of suboptimal model design in healthcare AI for better medical predictions, though it is incremental as it builds on existing neural architecture search techniques.
The paper tackled the challenge of designing deep learning models for multimodal Electronic Health Record (EHR) data by proposing AutoFM, a neural architecture search framework that automatically finds optimal architectures and fusion strategies, achieving significant performance improvements over state-of-the-art methods.
The widespread adoption of Electronic Health Record (EHR) systems in healthcare institutes has generated vast amounts of medical data, offering significant opportunities for improving healthcare services through deep learning techniques. However, the complex and diverse modalities and feature structures in real-world EHR data pose great challenges for deep learning model design. To address the multi-modality challenge in EHR data, current approaches primarily rely on hand-crafted model architectures based on intuition and empirical experiences, leading to sub-optimal model architectures and limited performance. Therefore, to automate the process of model design for mining EHR data, we propose a novel neural architecture search (NAS) framework named AutoFM, which can automatically search for the optimal model architectures for encoding diverse input modalities and fusion strategies. We conduct thorough experiments on real-world multi-modal EHR data and prediction tasks, and the results demonstrate that our framework not only achieves significant performance improvement over existing state-of-the-art methods but also discovers meaningful network architectures effectively.