DiReCT: Diagnostic Reasoning for Clinical Notes via Large Language Models

2024-08-04

arXiv.org

AI Summary

Existing large language models (LLMs) exhibit insufficient interpretability in clinical diagnostic reasoning, particularly failing to replicate the complete human physician reasoning chain—“observation → inference → diagnosis.” Method: We introduce DiReCT, the first diagnostic reasoning benchmark, comprising 511 physician-annotated clinical notes and an accompanying domain-specific medical knowledge graph. We propose a novel evaluation paradigm that jointly leverages structured knowledge graphs and unstructured clinical text for reasoning assessment. Contribution/Results: DiReCT is the first benchmark to systematically quantify LLM performance along two interpretability dimensions: diagnostic pathway plausibility and conclusion accuracy. Experiments reveal that even state-of-the-art models (e.g., GPT-4), when augmented with domain knowledge, significantly underperform human physicians in both dimensions. These findings underscore the critical need for knowledge-enhanced architectures and explicitly interpretable modeling in clinical AI systems.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Machine Learning: Large Multimodal Models (LMMs)
• Reasoning under Uncertainty: Other Foundations of Reasoning under Uncertainty

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large language models (LLMs) have recently showcased remarkable capabilities, spanning a wide range of tasks and applications, including those in the medical domain. Models like GPT-4 excel in medical question answering but may face challenges in the lack of interpretability when handling complex tasks in real clinical settings. We thus introduce the diagnostic reasoning dataset for clinical notes (DiReCT), aiming at evaluating the reasoning ability and interpretability of LLMs compared to human doctors. It contains 511 clinical notes, each meticulously annotated by physicians, detailing the diagnostic reasoning process from observations in a clinical note to the final diagnosis. Additionally, a diagnostic knowledge graph is provided to offer essential knowledge for reasoning, which may not be covered in the training data of existing LLMs. Evaluations of leading LLMs on DiReCT bring out a significant gap between their reasoning ability and that of human doctors, highlighting the critical need for models that can reason effectively in real-world clinical scenarios.

Problem

• Large Language Models
• Medical Reasoning
• Complex Healthcare Scenarios

Innovation

• DiReCT Dataset
• Knowledge Graph Assisted Models
• Medical Reasoning Evaluation