CSTRL: Context-Driven Sequential Transfer Learning for Abstractive Radiology Report Summarization

2025-02-21

arXiv.org

AI Summary

This study addresses the clinical need for automatically generating Impression sections from Findings in radiology reports, tackling challenges including medical terminology complexity, contextual drift, and stringent factual consistency requirements. We propose a context-driven sequential transfer learning framework that innovatively integrates Fisher information matrix regularization to mitigate parameter degradation and catastrophic forgetting, while jointly incorporating medical context modeling and factual consistency constraints to ensure accurate clinical information extraction and coherent diagnostic reasoning. Evaluated on MIMIC-CXR and Open-I datasets, our method achieves +56.2% improvement in BLEU-1 and +41.0% in ROUGE-2 over state-of-the-art approaches. Human evaluation further confirms its superior factual consistency and clinical context fidelity. The framework establishes a reliable, clinically grounded pathway for automated radiology report generation.

Technology Category

• Machine Learning: Multimodal Learning
• Natural Language Processing: Information Extraction
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

A radiology report comprises several sections, including the Findings and Impression of the diagnosis. Automatically generating the Impression from the Findings is crucial for reducing radiologists' workload and improving diagnostic accuracy. Pretrained models that excel in common abstractive summarization problems encounter challenges when applied to specialized medical domains largely due to the complex terminology and the necessity for accurate clinical context. Such tasks in medical domains demand extracting core information, avoiding context shifts, and maintaining proper flow. Misuse of medical terms can lead to drastic clinical errors. To address these issues, we introduce a sequential transfer learning that ensures key content extraction and coherent summarization. Sequential transfer learning often faces challenges like initial parameter decay and knowledge loss, which we resolve with the Fisher matrix regularization. Using MIMIC-CXR and Open-I datasets, our model, CSTRL - Context-driven Sequential TRansfer Learning - achieved state-of-the-art performance, showing 56.2% improvement in BLEU-1, 40.5% in BLEU-2, 84.3% in BLEU-3, 28.9% in ROUGE-1, 41.0% in ROUGE-2 and 26.5% in ROGUE-3 score over benchmark studies. We also analyze factual consistency scores while preserving the medical context. Our code is publicly available at https://github.com/fahmidahossain/Report_Summarization.

Problem

• Automatically generate radiology report summaries from findings
• Address challenges of medical terminology and clinical context
• Prevent knowledge loss in sequential transfer learning

Innovation

• Sequential transfer learning for medical summarization
• Fisher matrix regularization prevents knowledge loss
• Context-driven approach ensures clinical accuracy