A General Knowledge Injection Framework for ICD Coding

2025-05-24

AI Summary

To address the severe long-tail distribution of ICD codes, the absence of code-level evidence annotations, and the limited generalizability and scalability of existing methods—stemming from reliance on single knowledge sources and complex, task-specific modules—this paper proposes the General Knowledge Injection framework for ICD coding (GKI-ICD). GKI-ICD is the first approach to modularization-free, unified modeling of three heterogeneous knowledge types: ICD code descriptions, synonyms, and hierarchical relationships. It jointly encodes these knowledge sources via a knowledge-aware attention mechanism and a hierarchical encoder, preserving both their distinctiveness and complementarity. Evaluated on mainstream ICD coding benchmarks, GKI-ICD consistently outperforms state-of-the-art methods, achieving significant improvements across multiple metrics. The implementation is publicly available.

Technology Category

• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Data Mining & Knowledge Management: Knowledge Acquisition from the Web
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

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: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

ICD Coding aims to assign a wide range of medical codes to a medical text document, which is a popular and challenging task in the healthcare domain. To alleviate the problems of long-tail distribution and the lack of annotations of code-specific evidence, many previous works have proposed incorporating code knowledge to improve coding performance. However, existing methods often focus on a single type of knowledge and design specialized modules that are complex and incompatible with each other, thereby limiting their scalability and effectiveness. To address this issue, we propose GKI-ICD, a novel, general knowledge injection framework that integrates three key types of knowledge, namely ICD Description, ICD Synonym, and ICD Hierarchy, without specialized design of additional modules. The comprehensive utilization of the above knowledge, which exhibits both differences and complementarity, can effectively enhance the ICD coding performance. Extensive experiments on existing popular ICD coding benchmarks demonstrate the effectiveness of GKI-ICD, which achieves the state-of-the-art performance on most evaluation metrics. Code is available at https://github.com/xuzhang0112/GKI-ICD.

Problem

• Addresses long-tail distribution and lack of annotations in ICD coding
• Integrates multiple knowledge types without specialized modules
• Enhances ICD coding performance using complementary knowledge sources

Innovation

• Integrates ICD Description, Synonym, Hierarchy knowledge
• No specialized modules, general framework design
• Achieves state-of-the-art ICD coding performance