LCDL: Classification of ICD codes based on disease label co-occurrence dependency and LongFormer with medical knowledge

Yumeng Yang, Hongfei Lin, Zhihao Yang, Yijia Zhang, Di Zhao, Ling Luo

Abstract: Highlights•The proposed LongFormer model is grounded on the co-occurrence of disease labels, which aims to handle lengthy text records within electronic medical records while mitigating the long-tail phenomenon associated with disease labels.•We present an external knowledge base in the field of medicine to address the challenge of semantic ambiguity that arises from limited training data.•Our LCDL model underwent thorough evaluation using the MIMIC-III dataset, revealing that it can achieve greater success in classifying disease codes.