EUVIMEDEUVIMED entdecken
Open Access

Large language models for biomedicine: foundations, opportunities, challenges, and best practices

Satya S Sahoo iD, Joseph M Plasek iD, Hua Xu iD, Özlem Uzuner iD, Trevor Cohen, Meliha Yetisgen, Hongfang Liu, Stéphane Meystre iD, Yanshan Wang

Journal of the American Medical Informatics Association · 01.09.2024 · 31 (9) · 2114-2124

DOI: 10.1093/jamia/ocae074 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of the American Medical Informatics Association, online erschienen am 24.04.2024. Verfasst von Satya S Sahoo und 8 weiteren Autor:innen. Beteiligte Einrichtung: Department of Population and Quantitative Health Sciences, School of Medicine, Case Western Reserve University, Cleveland, OH 44122, United States. Hauptthemen laut MEDLINE: Natural Language Processing, Neural Networks, Computer. Die Arbeit wurde bisher 64-mal zitiert (Crossref) und nennt 56 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: US National Institutes of Health, US Department of Defense, Dravet Syndrome Foundation.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objectives Generative large language models (LLMs) are a subset of transformers-based neural network architecture models. LLMs have successfully leveraged a combination of an increased number of parameters, improvements in computational efficiency, and large pre-training datasets to perform a wide spectrum of natural language processing (NLP) tasks. Using a few examples (few-shot) or no examples (zero-shot) for prompt-tuning has enabled LLMs to achieve state-of-the-art performance in a broad range of NLP applications. This article by the American Medical Informatics Association (AMIA) NLP Working Group characterizes the opportunities, challenges, and best practices for our community to leverage and advance the integration of LLMs in downstream NLP applications effectively. This can be accomplished through a variety of approaches, including augmented prompting, instruction prompt tuning, and reinforcement learning from human feedback (RLHF). Target Audience Our focus is on making LLMs accessible to the broader biomedical informatics community, including clinicians and researchers who may be unfamiliar with NLP. Additionally, NLP practitioners may gain insight from the described best practices. Scope We focus on 3 broad categories of NLP tasks, namely natural language understanding, natural language inferencing, and natural language generation. We review the emerging trends in prompt tuning, instruction fine-tuning, and evaluation metrics used for LLMs while drawing attention to several issues that impact biomedical NLP applications, including falsehoods in generated text (confabulation/hallucinations), toxicity, and dataset contamination leading to overfitting. We also review potential approaches to address some of these current challenges in LLMs, such as chain of thought prompting, and the phenomena of emergent capabilities observed in LLMs that can be leveraged to address complex NLP challenge in biomedical applications.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Natural Language ProcessingNeural Networks, ComputerMedical InformaticsHumans

Stichwörter der Autor:innen

clinical natural language processinglarge language modelsmedical informatics applicationstransfer learningtransformer neural networks

Autor:innen und Institutionen

  • Department of Population and Quantitative Health Sciences, School of Medicine, Case Western Reserve University, Cleveland, OH 44122, United States.
  • Division of General Internal Medicine and Primary Care, Brigham and Women's Hospital, Harvard Medical School, Cambridge, MA 02115, United States.
  • Section of Biomedical Informatics and Data Science, School of Medicine, Yale University, New Haven, CT 06510, United States.
  • Department of Information Sciences and Technology, George Mason University, Fairfax, VA 22030, United States.
  • Department of Biomedical Informatics and Medical Education, University of Washington, Seattle, WA 98109, United States.
  • Department of Health Data Science and Artificial Intelligence, The University of Texas Health Science Center at Houston, Houston, TX 77030, United States.

ORCID-Profile: Satya S Sahoo, Joseph M Plasek, Hua Xu, Özlem Uzuner, Stéphane Meystre

Förderung

  • US National Institutes of Health
  • US Department of Defense
  • Dravet Syndrome Foundation
  • NCATS NIH HHS
  • NIH HHS
  • NIH
  • National Center for Advancing Translational Sciences, Clinical and Translational Science
  • NLM NIH HHS

Zitieren

APA

Sahoo, S. S., Plasek, J. M., Xu, H., Uzuner, Ö., Cohen, T., Yetisgen, M., Liu, H., Meystre, S., & Wang, Y. (2024). Large language models for biomedicine: foundations, opportunities, challenges, and best practices. Journal of the American Medical Informatics Association, 31(9), 2114-2124. https://doi.org/10.1093/jamia/ocae074

Vancouver

Sahoo SS, Plasek JM, Xu H, Uzuner Ö, Cohen T, Yetisgen M, et al. Large language models for biomedicine: foundations, opportunities, challenges, and best practices. Journal of the American Medical Informatics Association. 2024;31(9):2114-2124. doi:10.1093/jamia/ocae074

BibTeX
@article{Sahoo2024,
  author = {Sahoo, Satya S and Plasek, Joseph M and Xu, Hua and Uzuner, Özlem and Cohen, Trevor and Yetisgen, Meliha and Liu, Hongfang and Meystre, Stéphane and Wang, Yanshan},
  title = {Large language models for biomedicine: foundations, opportunities, challenges, and best practices},
  journal = {Journal of the American Medical Informatics Association},
  year = {2024},
  volume = {31},
  number = {9},
  pages = {2114--2124},
  doi = {10.1093/jamia/ocae074}
}
Kostenlos starten →