Automating Cancer Registries: Overcoming Data Bottlenecks with AI-Powered Abstraction

Cancer registries sit at the heart of oncology research and public health, capturing patient demographics, tumor characteristics, staging, treatment, biomarkers, and outcomes. Today this work is done by manual abstraction, where Certified Tumor Registrars (CTRs) spend hours per case reviewing unstructured pathology reports, clinical notes, and lab results. The result: years-long reporting delays and data that’s outdated by the time it’s used.

A cancer registry can receive 80,000 pathology reports from a single lab, only to realize that a mere 4% of those reports are relevant to actual patient cases. Filtering, coding, and validating this information under frameworks like SEER, AJCC, and NAACCR is a monumental task, demanding precision across thousands of decision rules and site-specific coding nuances. Traditional automation and general-purpose LLMs struggle with this complexity, often failing to meet regulatory-grade accuracy or handle multimodal data consistently.

This webinar will unpack the key challenges and practical solutions for scaling cancer registry automation, including:

• The Manual Bottleneck: How human-only workflows severely limit scalability and timeliness.
• Why Previous AI Approaches Fell Short: Limitations of LLM and NLP methods in achieving regulatory-grade data abstraction.
• Automation Requirements: Achieving compliance, precision, and longitudinal understanding of multimodal patient data within NAACCR and SEER frameworks.
• The Human-AI Loop: How expert validation complements automation to ensure ongoing data quality and audit readiness.
• John Snow Labs’ Approach: A hybrid end-to-end system combining healthcare-specific language models, oncology-specific agents, structured schema compliance, and human oversight.

Attendees will gain a practical framework for deploying AI in cancer registry workflows, accelerating data throughput while maintaining the transparency, accuracy, and compliance essential for real-world oncology research.