Enhancing Medication Recommendation with LLM Text Representation

NCU Institutional Repository > 管理學院 > 資訊管理研究所 > 博碩士論文 > Item 987654321/95555

jsp.display-item.identifier=請使用永久網址來引用或連結此文件: http://ir.lib.ncu.edu.tw/handle/987654321/95555

题名:	Enhancing Medication Recommendation with LLM Text Representation
作者:	李育慈;Lee, Yu-Tzu
贡献者:	資訊管理學系
关键词:	藥物推薦;電子醫療病例;臨床筆記;大型語言模型;知識提取;Medication Recommendation;EMR/EHR;Clinical Notes;Large Language Model;Knowledge Extraction
日期:	2024-07-26
上传时间:	2024-10-09 17:00:42 (UTC+8)
出版者:	國立中央大學
摘要:	大多數現有的藥物推薦模型僅使用結構化數據(如醫療代碼)進行預測，而對於大量的非結構化或半結構化數據利用不足。為了有效增加利用率，我們提出了一種通過大型語言模型(LLM)文本表徵增強藥物推薦的方法。LLM 具備強大的語言理解和生成能力，能夠從包含複雜術語的臨床筆記等複雜且冗長的非結構化數據中提取資訊。這種方法可以應用於我們選擇的幾個已被提出的模型，並通過文本和醫療代碼的組合表徵在兩個不同數據集上的實驗中提高藥物推薦性能:著名的醫療數據集 MIMIC-III 和嘉義基督教醫院(Ditmanson Medical Foundation Chia-Yi Christian Hospital; CYCH)的住院數據。實驗結果表明，LLM 文本表徵能夠提高我們選擇的大多數基礎模型在兩個數據集上的表現。僅使用 LLM 文本表徵甚至可以展示出與僅使用醫療代碼表徵相當的能力。總體而言，這是一種可以應用於其他模型並提高藥物推薦表現的通用方法。透過使用 LLM，我們減少了傳統方法中處理文本的冗餘。通過結合結構化和非結構化數據，我們優化了 EMR 數據的利用，解決了利用不足的問題。;Most of the existing medication recommendation models are predicted with only structured data such as medical codes, with the remaining other large amount of unstructured or semi-structured data underutilization. To increase the utilization effectively, we proposed a method of enhancing medication recommendation with Large Language Model (LLM) text representation. LLM harnesses powerful language understanding and generation capabilities, enabling the extraction of information from complex and lengthy unstructured data such as clinical notes which contain complex terminology. This method can be applied to several existing medication recommendation models we selected and improve medication recommendation performance with the combination representation of text and medical code experiments on two different datasets: the well-known medical datasets MIMI-III and hospitalized data from Ditmanson Medical Foundation Chia-Yi Christian Hospital (CYCH). The experiment results show that LLM text representation can improve most base models we selected on both datasets. LLM text representation alone can even demonstrate a comparable ability to the medical code representation alone. Overall, this is a general method that can be applied to other models and datasets for improved prediction performance. With LLM, we reduce the redundancy of the processing text in traditional approaches. By combining the structured and unstructured data, we optimize the utilization of EMR data, addressing the issue of underutilization.
显示于类别:	[資訊管理研究所] 博碩士論文

文件中的档案:

档案	描述	大小	格式	浏览次数
index.html		0Kb	HTML	41	检视/开启

在NCUIR中所有的数据项都受到原著作权保护.

社群 sharing

数据加载中.....