SentryLine helps doctors answer questions from changing cancer care guidelines
SentryLine: Evidence-Grounded Question Answering over Evolving Documents in Oncology Care
Computation and Language
Summary
Cancer treatment guidelines change often as new research comes out, making it hard for doctors to keep up. The researchers created SentryLine, a system that finds and explains answers from the latest cancer care guidelines. It shows where answers come from, checks facts and dates, and notes if the guidelines have recently changed. They tested SentryLine with real clinical questions and found it gave better answers than other tools, especially when complex reasoning or doctor-specific advice was needed.
OncologyClinical guidelinesQuestion answeringNatural language processingRetrieval augmented generationEvidence synthesisConversational AITemporal verificationRole-specific adaptationBenchmark evaluation
Authors
Tampu Ravi Kumar, Gaurav Najpande, Muhammad Ali Khan, Kaneez Zahra Rubab Khakwani, Karan Kathuria, Shorya Azriel Moses, Yuvraj Kalia, M Bassam Sonbol, Irbaz Bin Riaz, Vivek Gupta
Abstract
Oncology care operates at constant pressure of absorbing rapidly evolving evidence base in biomedicine. The American Society of Clinical Oncology (ASCO) addresses this through living guidelines, but the format introduces a new burden: any recommendation can change at any point, across multiple versioned documents. We present SENTRYLINE, a living guideline-aware clinical question answering system. SENTRYLINE retrieves guideline passages through a vectorless hierarchical RAG pipeline and returns a role-specific answer with inline citations, factual and temporal verification reports, and drift detection notes that surface when a guideline has been updated. We construct ASCOBENCH, a benchmark of 405 three-turn conversations across four question categories with gold answers from expert annotators(clinicians), and use test set to evaluate SENTRYLINE against five baselines under an LLM-as-judge framework. Experiments across three generation backbones show consistent improvements over four retrieval baselines and ASCO's guideline assistant, with particularly strong gains on Reasoning and Role-Specific questions where multi-hop synthesis and register adaptation are required