Related Experiment Video
Updated: Aug 6, 2026
05:19
A New Technique for Treating Low-risk Prostate Cancer—Super Active Surveillance
Published on: November 7, 2025
A Supervised Fine-Tuned Large Language Model for Lifestyle Management in Patients With Prostate Cancer: Development
Fangyuan Jiang1, Qiuwen Yang1, Xin Zheng1
1Department of Medical Informatics, School of Medicine, Nantong University, Qixiu Road 19#, Nantong, Jiangsu, 226001, China, +86 0513 8505 1891.
Journal of Medical Internet Research
|July 21, 2026
Summary
A new AI model, Prostate Cancer Patient Lifestyle Management Model via Supervised Fine-Tuning (PCaPLMM_SFT), offers personalized lifestyle advice for prostate cancer patients. This model improves health literacy and self-management, outperforming existing AI systems in evaluations.
Area of Science:
- Artificial Intelligence in Oncology
- Digital Health Interventions
- Patient Self-Management Support
Background:
- Prostate cancer patients benefit from lifestyle interventions for improved adherence and quality of life.
- Existing large language models (LLMs) lack the capability for safe, evidence-based, individualized lifestyle recommendations for these patients.
Purpose of the Study:
- To develop and evaluate a supervised fine-tuned LLM, PCaPLMM_SFT, for prostate cancer patients.
- To enhance patient health literacy and support lifestyle self-management.
Main Methods:
- A structured lifestyle management knowledge base was created from literature (PubMed, Feb 2015-Feb 2025).
- Retrieval-augmented generation was used to create bilingual English-Chinese question-answer pairs.
- The model (Baichuan2-7B-Chat) underwent supervised fine-tuning and was evaluated by referee LLMs and domain experts.
Main Results:
- The PCaPLMM_SFT-Train dataset was constructed from 2211 publications, yielding over 150,000 knowledge slices.
- PCaPLMM_SFT demonstrated superior performance compared to Baichuan2-7B-Chat and comparable or better results than GPT-3.5-Turbo.
- Evaluation showed robust and consistent performance across multiple lifestyle scenarios.
Conclusions:
- PCaPLMM_SFT proves the feasibility of creating medical LLMs using structured knowledge and QA data.
- The developed framework offers a reproducible method for evidence-based health education and lifestyle management.
- This research lays the foundation for real-world health management applications.