Researchers have developed a new framework called Semantic Medical Identifier (SMI) to improve the representation of medical concepts in language models. SMI addresses the limitation of current models that often fail to capture the hierarchical structure among medical terms, particularly for rare or fine-grained concepts. By grounding Semantic IDs in expert-defined medical ontologies, SMI aggregates embeddings from frozen biomedical language models to create representations that preserve coarse-to-fine clinical semantics. Experiments on clinical datasets like MIMIC-IV and eICU demonstrate that SMI enhances performance in tasks such as mortality prediction, readmission prediction, and drug recommendation, while also yielding more robust patient representations under distribution shifts. AI
IMPACT Enhances clinical task performance by improving the semantic understanding of medical concepts in language models.
RANK_REASON The cluster contains an academic paper detailing a new framework for medical concept representation. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
