ICDXML: enhancing ICD coding with probabilistic label trees and dynamic semantic representations

• Published in: Scientific reports 2024-08, Vol.14 (1), p.18319-12, Article 18319
• Main Authors: Wang, Zeqiang, Wang, Yuqi, Zhang, Haiyang, Wang, Wei, Qi, Jun, Chen, Jianjun, Sastry, Nishanth, Johnson, Jon, De, Suparna
• Format: Article
• Language: English
• Subjects: 639/705/1042; 639/705/117; Algorithms; Automation; Classification; Computer science; Databases, Factual; Electronic health records; Extreme multi-label classification; Few-shot learning; Humanities and Social Sciences; Humans; ICD coding; International Classification of Diseases; Knowledge; Language; Medical knowledge representation; Medical research; multidisciplinary; Natural Language Processing; Neural coding; Neural networks; Neural Networks, Computer; Science; Science (multidisciplinary); Semantics; Taxonomy
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-024-69214-9

Accurately assigning standardized diagnosis and procedure codes from clinical text is crucial for healthcare applications. However, this remains challenging due to the complexity of medical language. This paper proposes a novel model that incorporates extreme multi-label classification tasks to enhance International Classification of Diseases (ICD) coding. The model utilizes deformable convolutional neural networks to fuse representations from hidden layer outputs of pre-trained language models and external medical knowledge embeddings fused using a multimodal approach to provide rich semantic encodings for each code. A probabilistic label tree is constructed based on the hierarchical structure existing in ICD labels to incorporate ontological relationships between ICD codes and enable structured output prediction. Experiments on medical code prediction on the MIMIC-III database demonstrate competitive performance, highlighting the benefits of this technique for robust clinical code assignment.