FT-MDT: Extracting Decision Trees from Medical Texts via a Novel Low-rank Adaptation Method

2025-10-06

AI Summary

Medical decision tree (MDT) construction traditionally relies heavily on labor-intensive manual annotation, hindering the development of clinical decision support systems. To address this, we propose PI-LoRA—a novel low-rank adaptation method that incorporates path-integrated gradient information to explicitly model inter-module synergistic effects, enabling dynamic rank allocation and redundant parameter pruning. PI-LoRA preserves a lightweight architecture while significantly enhancing the accuracy and robustness of medical knowledge extraction from text. Evaluated on multiple authoritative clinical guideline datasets, PI-LoRA achieves state-of-the-art performance with substantially reduced model complexity—requiring 37%–52% fewer parameters—and yields an average 4.8-point improvement in F1 score. This work establishes an efficient, reliable end-to-end paradigm for automated, text-to-MDT generation.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Natural Language Processing: Information Extraction
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

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

Abstract

Knowledge of the medical decision process, which can be modeled as medical decision trees (MDTs), is critical to building clinical decision support systems. However, current MDT construction methods rely heavily on time-consuming and laborious manual annotation. To address this challenge, we propose PI-LoRA (Path-Integrated LoRA), a novel low-rank adaptation method for automatically extracting MDTs from clinical guidelines and textbooks. We integrate gradient path information to capture synergistic effects between different modules, enabling more effective and reliable rank allocation. This framework ensures that the most critical modules receive appropriate rank allocations while less important ones are pruned, resulting in a more efficient and accurate model for extracting medical decision trees from clinical texts. Extensive experiments on medical guideline datasets demonstrate that our PI-LoRA method significantly outperforms existing parameter-efficient fine-tuning approaches for the Text2MDT task, achieving better accuracy with substantially reduced model complexity. The proposed method achieves state-of-the-art results while maintaining a lightweight architecture, making it particularly suitable for clinical decision support systems where computational resources may be limited.

Problem

• Automatically extracting medical decision trees from clinical texts
• Reducing manual annotation workload in medical decision modeling
• Improving accuracy while maintaining lightweight model architecture

Innovation

• Novel low-rank adaptation method for MDT extraction
• Integrates gradient path for synergistic module effects
• Prunes less important modules for efficient model