aiqbench analysis
This research introduces Hyperbolic Clinical Ontology Embeddings (HCOE), a method that maps frozen BioBERT embeddings into a Poincaré ball to explicitly preserve medical code hierarchies like ICD and ATC. By integrating ontology-guided contrastive learning, the approach significantly outperforms standard models on clinical relation prediction and downstream tasks such as mortality and medication recommendation within the MIMIC-IV dataset. The work addresses a critical gap in biomedical language modeling where textual semantics often fail to capture structured hierarchical relationships essential for clinical decision support.
HCOE: Hyperbolic Clinical Ontology Embeddings from Biomedical Language Models
arXiv:2609.30763v1 Announce Type: cross Abstract: Biomedical language models (LMs) encode textual semantics but do not explicitly preserve medical code hierarchies. We present Hyperbolic Clinical Ontology Embeddings (HCOE) for hierarchy-aware clinical concept representation. HCOE maps frozen BioBERT embeddings into a Poincare ball, combining parent-side and child-side ontology-guided contrastive learning with coarse-to-fine ontology-path aggregation. It uses International Classification of Diseases (ICD) codes organized by Clinical Classifications Software (CCS) and Anatomical Therapeutic Chemical (ATC) medication hierarchies. Evaluations show that HCOE performs best on ICD/ATC clinical relation prediction and CCS-to-PheCode hierarchy transfer. On the MIMIC-IV dataset, HCOE also achieves the best performance on mortality prediction, readmission prediction, medication recommendation, and rare drug prediction.
Abstract:Biomedical language models (LMs) encode textual semantics but do not explicitly preserve medical code hierarchies. We present Hyperbolic Clinical Ontology Embeddings (HCOE) for hierarchy-aware clinical concept representation. HCOE maps frozen BioBERT embeddings into a Poincare ball, combining parent-side and child-side ontology-guided contrastive learning with coarse-to-fine ontology-path aggregation. It uses International Classification of Diseases (ICD) codes organized by Clinical Classifications Software (CCS) and Anatomical Therapeutic Chemical (ATC) medication hierarchies. Evaluations show that HCOE performs best on ICD/ATC clinical relation prediction and CCS-to-PheCode hierarchy transfer. On the MIMIC-IV dataset, HCOE also achieves the best performance on mortality prediction, readmission prediction, medication recommendation, and rare drug prediction.
Accepted at IEEE BIBM 2026. 7 pages, 3 figures, 4 tables
Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
arXiv:2609.30763 [cs.AI]
(or arXiv:2609.30763v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.30763 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 212 words
Related coverage
SMI: Ontology-Grounded Semantic IDs for Hierarchy-Aware Medical Concept Representation
arXiv · Oct 15, 2026
EHRAdapt: Adapting Pretrained Language Models to Electronic Health Records with Semantic Priors for Rare Clinical Events
arXiv · Sep 29, 2026
SNOMED CT Concept Recommendation from Masked Clinical Context
arXiv · Sep 17, 2026
Neighborhood-Aware Dual Biomedical Entity Linking
arXiv · Aug 6, 2026