Automatic rare disease extraction based on large language models

Cao, Lang

Permalink

https://hdl.handle.net/2142/124192 Copy

Description

Title

Automatic rare disease extraction based on large language models

Author(s)

Cao, Lang

Issue Date

2024-04-05

Director of Research (if dissertation) or Advisor (if thesis)

Sun, Jimeng

Department of Study

Computer Science

Discipline

Computer Science

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• AI for Healthcare
• Large Language Model
• Natural Language Processing

Abstract

Identifying and extracting information on rare diseases is crucial in various medical contexts. However, mature rare disease extraction methods are lacking in low-resource settings. In this paper, we aim to create an end-to-end system called AutoRD, which automates extracting information from clinical text about rare diseases. We achieve this using large language models and medical knowledge graphs developed from open-source medical ontologies. Large language models (LLMs) aid in language analysis, while knowledge graphs provide content-specific facts, thus filling in any information gaps. Our system, AutoRD, is a software pipeline involving data preprocessing, entity extraction, relation extraction, entity calibration, and knowledge graph construction. Large language models and open-source data are leveraged throughout the pipeline. We have conducted various tests to evaluate the performance of AutoRD and highlighted its strengths and limitations in this paper. We quantitatively evaluate our system in terms of entity extraction, relation extraction, and the performance of knowledge graph construction. AutoRD achieves an overall F1 score of 47.3%, an improvement of 0.8% compared to the fine-tuned model, and a 14.4% improvement compared to the base LLM. In detail, AutoRD achieves an overall entity extraction F1 score of 56.1% (rare_disease: 83.5%, disease: 35.8%, symptom_and_sign: 46.1%, anaphor: 67.5%) and an overall relation extraction F1 score of 38.6% (produces: 34.7%, increases_risk_of: 12.4%, is_a: 37.4%, is_acronym: 44.1%, is_synonym: 16.3%, anaphora: 57.5%). Our qualitative experiment also demonstrates that the performance in constructing the knowledge graph is commendable. Several designs, including the incorporation of ontologies-enhanced LLMs, contribute to the improvement of AutoRD. AutoRD demonstrates superior performance compared to other methods, demonstrating the potential of LLM applications in rare disease detection and AI for healthcare.

Graduation Semester

2024-05

Type of Resource

Thesis

Copyright and License Information

Copyright 2024 Lang Cao

Owning Collections