CHRONOS: Extracting Novel Spine Surgery Hypotheses from Historical Medical Texts

This paper presents Chronos, a system for representing and revisiting overlooked historical and traditional medical insights using modern knowledge frameworks. Inspired by Nobel laureate Tu Youyou's systematic approach to mining traditional Chinese medicine for antimalarial compounds, Chronos employs large language models and decentralized knowledge graphs to formalize historical spine surgery observations into testable hypotheses. We implement the Hypothesis and Evidence (HypE) taxonomy to structure machine-readable, refutable hypotheses from historical and traditional medical texts, bridging centuries of medical knowledge. Using a comprehensive taxonomy spanning historical observations, formalized hypotheses, evidence elements, modern medical concepts, and research opportunities, we demonstrate Chronos' capabilities by generating novel research hypotheses from two sources: the works of Charles-Prosper Ollivier d'Angers (1824) and a 19th-century Thai traditional medicine compendium. These hypotheses—ranging from spinal venous congestion to gut-spine axis mechanisms—are evaluated for scientific merit and novelty, illustrating how AI-powered knowledge mining can uncover valuable insights from diverse historical medical literature. The Chronos system represents a scalable, automated approach to hypothesis generation with the potential to accelerate discovery across various medical domains.