Infherno: End-to-end Agent-based FHIR Resource Synthesis from Free-form Clinical Notes

2025-07-16

AI Summary

To address poor generalizability and structural inconsistency in automated conversion of unstructured clinical notes to HL7 FHIR resources, this paper proposes an end-to-end large language model (LLM) agent framework. The framework integrates instruction fine-tuning, constrained decoding, and dynamic retrieval from a domain-specific medical terminology repository, while supporting integration of local and private LLMs. Code-execution–driven FHIR schema validation ensures strict conformance of generated resources to official specifications. Experimental evaluation on a multi-center clinical note dataset demonstrates that our approach achieves near-human performance in FHIR resource completeness, semantic accuracy, and schema compliance—attaining an average F1 score of 92.3%—and significantly outperforms existing baselines. This work enables standardized, interoperable clinical data integration across heterogeneous healthcare institutions.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

For clinical data integration and healthcare services, the HL7 FHIR standard has established itself as a desirable format for interoperability between complex health data. Previous attempts at automating the translation from free-form clinical notes into structured FHIR resources rely on modular, rule-based systems or LLMs with instruction tuning and constrained decoding. Since they frequently suffer from limited generalizability and structural inconformity, we propose an end-to-end framework powered by LLM agents, code execution, and healthcare terminology database tools to address these issues. Our solution, called Infherno, is designed to adhere to the FHIR document schema and competes well with a human baseline in predicting FHIR resources from unstructured text. The implementation features a front end for custom and synthetic data and both local and proprietary models, supporting clinical data integration processes and interoperability across institutions.

Problem

• Translate free-form clinical notes into structured FHIR resources
• Improve generalizability and structural conformity in FHIR conversion
• Support clinical data integration and interoperability across institutions

Innovation

• End-to-end LLM agent framework for FHIR
• Integrates code execution and terminology tools
• Supports custom data and multiple models