VISTA Architect: A graph database-oriented health AI system demonstrated in multidisciplinary tumor boards

VISTA Architect is a new AI architecture designed for integrating large language models (LLMs) with longitudinal electronic health records (EHRs), transforming clinical documentation into a provenance-linked knowledge graph. It features a dual-layer structure: a source-faithful MEDS Graph for preserving EHR granularity and a Timeline Object Architecture (TOA) that uses graph-guided LLM extraction to synthesize patient histories, achieving 96.4% accuracy across critical tumor board variables in a study of 1,180 patients. This approach addresses limitations of long-context prompting and retrieval-augmented generation (RAG) by enabling efficient, organized patient data access and reducing preparation time significantly, making it relevant for practitioners aiming to enhance clinical data processing and accuracy in diverse medical specialties.