Frontiers in Artificial Intelligence (Jan 2025)
Fine-tuning a local LLaMA-3 large language model for automated privacy-preserving physician letter generation in radiation oncology
- Yihao Hou,
- Yihao Hou,
- Christoph Bert,
- Christoph Bert,
- Christoph Bert,
- Ahmed Gomaa,
- Ahmed Gomaa,
- Ahmed Gomaa,
- Godehard Lahmer,
- Godehard Lahmer,
- Godehard Lahmer,
- Daniel Höfler,
- Daniel Höfler,
- Daniel Höfler,
- Thomas Weissmann,
- Thomas Weissmann,
- Thomas Weissmann,
- Raphaela Voigt,
- Raphaela Voigt,
- Raphaela Voigt,
- Philipp Schubert,
- Philipp Schubert,
- Philipp Schubert,
- Charlotte Schmitter,
- Charlotte Schmitter,
- Charlotte Schmitter,
- Alina Depardon,
- Alina Depardon,
- Alina Depardon,
- Sabine Semrau,
- Sabine Semrau,
- Sabine Semrau,
- Andreas Maier,
- Rainer Fietkau,
- Rainer Fietkau,
- Rainer Fietkau,
- Yixing Huang,
- Yixing Huang,
- Florian Putz,
- Florian Putz,
- Florian Putz
Affiliations
- Yihao Hou
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Yihao Hou
- Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Christoph Bert
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Christoph Bert
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Christoph Bert
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Ahmed Gomaa
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Ahmed Gomaa
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Ahmed Gomaa
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Godehard Lahmer
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Godehard Lahmer
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Godehard Lahmer
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Daniel Höfler
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Daniel Höfler
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Daniel Höfler
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Thomas Weissmann
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Thomas Weissmann
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Thomas Weissmann
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Raphaela Voigt
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Raphaela Voigt
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Raphaela Voigt
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Philipp Schubert
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Philipp Schubert
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Philipp Schubert
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Charlotte Schmitter
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Charlotte Schmitter
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Charlotte Schmitter
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Alina Depardon
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Alina Depardon
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Alina Depardon
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Sabine Semrau
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Sabine Semrau
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Sabine Semrau
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Andreas Maier
- Pattern Recognition Lab, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Rainer Fietkau
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Rainer Fietkau
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Rainer Fietkau
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- Yixing Huang
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Yixing Huang
- Institute of Medical Technology, Health Science Center, Peking University, Beijing, China
- Florian Putz
- Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
- Florian Putz
- Comprehensive Cancer Center Erlangen-EMN, Erlangen, Germany
- Florian Putz
- Bavarian Cancer Research Center (BZKF), Erlangen, Germany
- DOI
- https://doi.org/10.3389/frai.2024.1493716
- Journal volume & issue
-
Vol. 7
Abstract
IntroductionGenerating physician letters is a time-consuming task in daily clinical practice.MethodsThis study investigates local fine-tuning of large language models (LLMs), specifically LLaMA models, for physician letter generation in a privacy-preserving manner within the field of radiation oncology.ResultsOur findings demonstrate that base LLaMA models, without fine-tuning, are inadequate for effectively generating physician letters. The QLoRA algorithm provides an efficient method for local intra-institutional fine-tuning of LLMs with limited computational resources (i.e., a single 48 GB GPU workstation within the hospital). The fine-tuned LLM successfully learns radiation oncology-specific information and generates physician letters in an institution-specific style. ROUGE scores of the generated summary reports highlight the superiority of the 8B LLaMA-3 model over the 13B LLaMA-2 model. Further multidimensional physician evaluations of 10 cases reveal that, although the fine-tuned LLaMA-3 model has limited capacity to generate content beyond the provided input data, it successfully generates salutations, diagnoses and treatment histories, recommendations for further treatment, and planned schedules. Overall, clinical benefit was rated highly by the clinical experts (average score of 3.4 on a 4-point scale).DiscussionWith careful physician review and correction, automated LLM-based physician letter generation has significant practical value.
Keywords