Scholarlink

The use of electronic health records (EHRs) involves considerable amounts of unstructured free text containing vital information related to clinical healthcare. However, these unstructured texts have proven to be hindrances in the process of automating healthcare systems because of the linguistic variability and lack of correspondence to standardized ICD-10 codes. To bridge this gap, a cost-effective and modular clinical pipeline has been proposed in this paper. The focus of this pipeline is to provide diagnosis and generate relevant ICD-10 codes. The clinical pipeline utilizes medical concept normalization of clinical healthcare terms based on medical knowledge and ICD-10 codes. Moreover, it involves a clinical reasoning process based on large language models and a knowledge graph driven search of clinical ICD-10 codes. The working of the pipeline involves normalizing clinical terms related to healthcare and transforms them into a structured format based on clinical knowledge. The clinical terms are transformed into a knowledge-driven narrative format based on reasoning and ICD-10 codes. The pipeline aims to provide vital clinical interpretation along with the generation of ICD-10 codes.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Zero-Cost Clinical Natural Language Processing Pipeline for International Classification of Diseases-10 Code Generation Using Knowledge-Based Prompt Engineering

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • E. Ajitha,
  • P. A. Alfina Valeri,
  • T. Irene

摘要

The use of electronic health records (EHRs) involves considerable amounts of unstructured free text containing vital information related to clinical healthcare. However, these unstructured texts have proven to be hindrances in the process of automating healthcare systems because of the linguistic variability and lack of correspondence to standardized ICD-10 codes. To bridge this gap, a cost-effective and modular clinical pipeline has been proposed in this paper. The focus of this pipeline is to provide diagnosis and generate relevant ICD-10 codes. The clinical pipeline utilizes medical concept normalization of clinical healthcare terms based on medical knowledge and ICD-10 codes. Moreover, it involves a clinical reasoning process based on large language models and a knowledge graph driven search of clinical ICD-10 codes. The working of the pipeline involves normalizing clinical terms related to healthcare and transforms them into a structured format based on clinical knowledge. The clinical terms are transformed into a knowledge-driven narrative format based on reasoning and ICD-10 codes. The pipeline aims to provide vital clinical interpretation along with the generation of ICD-10 codes.