John Snow Labs 63d ago

Cancer registry data is critical for research — but today it arrives 12–24 months too late to be useful in practice. This session shows how regulatory-grade AI turns manual cancer registry abstraction into a scalable, audit-ready system with full traceability. Dia Trambitas (Head of Product, John Snow Labs) breaks down how automated cancer registry abstraction can reach expert-level accuracy while remaining fully auditable and compliant. Instead of relying on generic LLM approaches, this system is built specifically for oncology workflows — combining multimodal clinical data, deterministic logic, and agentic reasoning to handle real-world complexity. The result: near real-time registry data, full provenance for every extracted field, and a system that supports both clinical research and regulatory requirements. From pathology reports and radiology data to longitudinal patient timelines, the platform reconstructs patient journeys and resolves conflicts that traditional document-level AI cannot handle. 📌 Applied Healthcare AI Summit 2026 — what actually works in real-world healthcare AI, from pilots to production systems. ⏱️ Key moments: 00:00 What cancer registries are and why they matter 03:13 Why registry data is delayed (manual bottlenecks) 04:40 Why generic LLMs fail in clinical workflows 07:32 Multimodal AI and patient-level reasoning 16:51 From manual abstraction to audit-ready, real-time registries #CancerRegistry #HealthcareAI #ClinicalData #RealWorldEvidence #AIinHealthcare #MedicalAI #DataProvenance #AIGovernance #OncologyData #ClinicalResearch