Guideline decision graph boosts radiology case accuracy and speed
Radiology Guideline-derived Decision Graph for Efficient Clinical Decision Support
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Medical guidelines for interpreting imaging tests can be complicated and scattered across many sections and tables. The authors created a system called GDGraph that breaks down these guidelines into clear decision steps. This system helps computers follow the guidelines when making recommendations, making their answers more accurate and easier to understand. Tests showed GDGraph outperforms other AI methods in interpreting kidney, ovary, and thyroid imaging cases quickly and correctly.
What this means in practice
- •For radiology departments: Automatically interpret complex imaging guidelines to improve diagnostic accuracy and reduce manual review time for kidney, ovarian, and thyroid cases.
- •For medical imaging software developers: Integrate a decision graph framework that generates traceable, guideline-compliant recommendations without task-specific model training, cutting costs and latency.
- •For clinical decision support product teams: Develop advanced decision support tools for radiologists using an automated guideline-driven decision graph that improves accuracy and transparency versus existing AI models.$Commercial implications: Enables building and selling diagnostic support software that more reliably follows official radiology guidelines, appealing to hospitals seeking better compliance.
Authors
Xuewen Zhou, Zhen Chen
Abstract
Radiological risk stratification and management planning are essential downstream tasks in medical imaging. Automating these tasks is challenging because clinical guideline criteria are distributed across prose, tables, and cross-referenced definitions. We present Guideline-derived Decision Graph (GDGraph), a framework that decouples guideline compilation from case evaluation. During compilation, a large language model partitions guidelines into section-based chunks, extracts local decision subgraphs with linked definitions, and aggregates related branches into an executable decision graph. During inference, a fast System One model answers localized clinical questions while deterministic traversal resolves Boolean logic, turning opaque single-pass predictions into a traceable decision path without task-specific training. Across 347 cases spanning renal cyst, adnexal mass, and thyroid nodule guidelines, GDGraph improves System One accuracy from 57.3% to 90.8%, outperforming retrieval-augmented GPT-4o (80.1%) and graph-guided GPT-4o with chain-of-thought (82.7%) at substantially lower latency and cost. The source code will be available.