IP-CRR: Information Pursuit for Interpretable Classification of Chest Radiology Reports

2025-04-30

AI Summary

To address insufficient interpretability and low clinical trust in AI-based chest X-ray report classification, this paper proposes an endogenous interpretable diagnostic model grounded in the principle of “explanation-as-mechanism.” Methodologically, it leverages Information Pursuit theory to identify highly discriminative clinical queries, integrating a Flan-T5–driven fact-checking module with a lightweight classifier to jointly generate disease classifications and query-answer–style explanations. Notably, this work pioneers the application of Information Pursuit to structured explanation generation for radiology reports. Evaluated on the MIMIC-CXR dataset, the model achieves high classification accuracy while producing clinically intelligible, traceable explanation chains—enhancing radiologists’ understanding of and confidence in AI decisions. This approach establishes a novel paradigm for deploying interpretable AI in radiology practice.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Computer Vision: Interpretability, Explainability, and Transparency
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

The development of AI-based methods for analyzing radiology reports could lead to significant advances in medical diagnosis--from improving diagnostic accuracy to enhancing efficiency and reducing workload. However, the lack of interpretability in these methods has hindered their adoption in clinical settings. In this paper, we propose an interpretable-by-design framework for classifying radiology reports. The key idea is to extract a set of most informative queries from a large set of reports and use these queries and their corresponding answers to predict a diagnosis. Thus, the explanation for a prediction is, by construction, the set of selected queries and answers. We use the Information Pursuit framework to select informative queries, the Flan-T5 model to determine if facts are present in the report, and a classifier to predict the disease. Experiments on the MIMIC-CXR dataset demonstrate the effectiveness of the proposed method, highlighting its potential to enhance trust and usability in medical AI.

Problem

• Develop interpretable AI for chest radiology report classification
• Address lack of interpretability in medical AI diagnostics
• Enhance trust in AI predictions via query-based explanations

Innovation

• Interpretable-by-design framework for radiology reports
• Information Pursuit selects most informative queries
• Flan-T5 model detects facts for diagnosis