使用统一多模态数据嵌入和模态感知注意力学习缺失模态的电子健康记录（英文版）

英文标题：Learning Missing Modal Electronic Health Records with Unified Multi-modal Data Embedding and Modality-Aware Attention中文摘要：本研究提出一种统一的多模态集合嵌入和具有跳过瓶颈的模态感知注意力（MAA），在 MIMIC-IV 数据集上预测死亡率、血管加压素需求和气管插管需求中，该模型优于其他基线模型，无需单独的补偿模块或易错的向前传递。英文摘要：Electronic Health Record (EHR) provides abundant information through variousmodalities. However, learning multi-modal EHR is currently facing two majorchallenges, namely, 1) data embedding and 2) cases with missing modality. Alack of shared embedding function ac