An agentic system for rare disease diagnosis with traceable reasoning - PubMed
5 days ago
- #healthcare-technology
- #rare-disease
- #AI-diagnosis
- DeepRare is a multi-agent system for rare disease diagnosis powered by large language models.
- It integrates over 40 specialized tools and up-to-date knowledge sources for differential diagnosis.
- The system processes heterogeneous clinical inputs, including free-text descriptions and genetic testing results.
- DeepRare generates ranked diagnostic hypotheses with transparent reasoning linked to medical evidence.
- It achieves an average Recall@1 of 57.18% in human-phenotype-ontology-based tasks, outperforming other methods by 23.79%.
- In multi-modal tests, DeepRare reaches 69.1% accuracy compared to Exomiser's 55.9% on 168 cases.
- Expert review confirmed 95.4% agreement on the validity and traceability of DeepRare's reasoning chains.
- The system aims to reduce the prolonged 'diagnostic odyssey' faced by rare disease patients.