Medical Knowledge-Enhanced ICD Coding Prediction

Fanfan Zhu

doi:10.1016/j.eswa.2026.132312

Medical Knowledge-Enhanced ICD Coding Prediction

2026-04-02

0

PRE

AI

DOI：10.1016/j.eswa.2026.132312

Original

Original request for help

Save

Abstract

En 中文

• Proposes a domain knowledge-enhanced pre-training framework for ICD coding, tackling concept recognition and semantic alignment challenges. • Develops a medical pre-training corpus integrating UMLS knowledge and entity linking for effective knowledge injection. • Employs dual-task learning (ETP & MLM) to deepen the model’s understanding of medical semantics. • Aims to improve medical concept standardization and fine-grained text-to-code semantic alignment.

Keywords:

ICD coding

medical knowledge

pre-training framework

semantic alignment

entity linking

Medical Knowledge-Enhanced ICD Coding Prediction

Abstract

Journal

Organization

Cited Papers