A specialized reasoning large language model for accelerating rare disease diagnosis: a randomized AI physician assistance trial

2026-06-23

AI Summary

This work addresses the diagnostic delays in rare diseases caused by the scarcity of specialized expertise and limitations of existing large language models in clinical deployability, evidence-based grounding, and data scale. The authors propose RaDaR (Rare Disease navigatoR), a 32-billion-parameter open-source reasoning-oriented large language model, which innovatively integrates 49,170 real-world cases with 104,666 phenotype-anchored narrative-generated synthetic cases through a reasoning-augmented training strategy. RaDaR achieves state-of-the-art performance across multiple public benchmarks and four external medical centers: in retrospective cohorts, it identifies the final diagnosis earlier in 61.06% of cases, reducing time to diagnosis by an average of 1.87 months; a randomized controlled trial demonstrates a 21.44-percentage-point improvement in diagnostic accuracy when assisting physicians. The study also establishes a reproducible framework for developing and validating diagnostic AI systems.

Abstract

Rare diseases affect millions of individuals worldwide, yet timely diagnosis remains a major public health challenge due to scarcity of specialized clinical expertise. While large language models (LLMs) show promise to support rare disease diagnosis, current models are constrained by insufficient clinical deployability, limited clinically grounded evidence, and scarcity of training data. Here we present RaDaR (Rare Disease navigatoR), an open-source, compact reasoning LLM (32B parameters) for rare disease diagnosis. RaDaR was trained with 49,170 publicly available free-text cases and 104,666 synthetic cases with reasoning-enhanced training. RaDaR showed the strongest performance among evaluated open-source models, including the 671B DeepSeek-R1, across public benchmarks and four external validation centers. In a retrospective cohort, RaDaR prioritized the final diagnosis before documented clinical suspicion in 61.06 percent of cases, corresponding to a potential lead time of 1.87 months and 50.18 percent of the within-center interval. In a randomized physician-assistance trial, RaDaR assistance improved physicians' rare-disease diagnostic accuracy by 21.44 percentage points compared with internet search alone. Synthetic-data ablations suggested that phenotype-anchored narratives provide useful training signal for long-tail rare diseases, with a monotonic scaling trend within the tested data range. Together, RaDaR and its development and validation framework provide a deployable rare-disease reasoning model and a reproducible development framework for diagnostic AI under data scarcity.

Problem

• rare disease diagnosis
• clinical deployability
• data scarcity
• diagnostic accuracy
• large language models

Innovation

• reasoning-enhanced LLM
• synthetic clinical data
• rare disease diagnosis
• phenotype-anchored narratives
• AI physician assistance trial