Search papers, labs, and topics across Lattice.
This study introduces ACIE (Agentic Clinical Information Extraction), an on-premise retrieval-augmented generation (RAG) pipeline designed to effectively extract and verify clinical information from extensive patient contexts. By addressing the challenges of temporal reasoning and cross-document dependencies, ACIE successfully grounds its outputs in source passages, achieving a high acceptance rate of 96.5% for extracted values verified by nuclear-medicine physicians. The findings highlight the critical importance of addressing metadata gaps in clinical AI applications, demonstrating that a tailored approach can significantly enhance the reliability of information extraction in healthcare settings.
Clinicians accepted 96.5% of extracted clinical data, showcasing a breakthrough in reliable information extraction from complex patient contexts.
Patient contexts span hundreds of heterogeneous documents and thousands of structured data points, yet the document-level metadata that AI systems need for retrieval and triage is absent or incomplete. Standard retrieval-augmented generation fails on this data, mishandling temporal reasoning, cross-document dependencies, and missing metadata. We deploy ACIE (Agentic Clinical Information Extraction) at University Medicine Essen: an on-premise agentic RAG pipeline that reasons over complete patient contexts and grounds every answer in source passages for clinician verification. We quantify the metadata gap, trace the architectural decisions it shaped, and evaluate extraction alongside an independent retrospective lymphoma registry study, in which nuclear-medicine physicians verify every extracted value against its cited sources. Across 7,326 judgments, clinicians accepted 96.5\% of extractions, with per-type acceptance ranging from 80\% to 99\%.
