The goal of this observational study is to develop and evaluate a large language model (LLM)-based decision support system for exercise prescription in breast cancer patients, aiming to provide personalized decision-making support for postoperative breast cancer rehabilitation.
The main questions it aims to answer are:
How accurate, personalized, and safe are the exercise prescriptions generated by the fine-tuned LLM? How does the model's performance compare with other mainstream or non-fine-tuned models across different stages and subtypes of breast cancer? Participants are postoperative breast cancer rehabilitation patients treated at Sun Yat-sen Memorial Hospital of Sun Yat-sen University. They will have demographic, tumor, treatment, and physical fitness data collected; receive personalized exercise prescriptions automatically generated by the LLM-based system; and provide subjective evaluations on the feasibility and executability of the prescriptions.
Adult patients aged 18-75 years with early-stage breast cancer who have undergone surgical treatment, such as mastectomy or breast-conserving surgery.
The patients had clear clinical diagnosis and complete electronic medical record information (including demographic information, tumor stage and classification, treatment history, physical performance evaluation data, etc.).
Presence of severe postoperative complications or comorbidities (e.g., uncontrolled cardiac or pulmonary disease) that may interfere with participation in rehabilitation or pose a safety risk.
Patients with breast cancer who have completed primary surgery and entered the postoperative rehabilitation stage at Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Guangzhou, China).All participants receive individualized exercise prescriptions generated by large language models under physician supervision and approval, and their feedback on the feasibility of these prescriptions is collected.
Cohort
Prospective
Not listed
Overall Quality Score of Exercise Prescriptions Based on a Five-Dimensional Expert Evaluation Framework
Each exercise prescription will be independently evaluated by six multidisciplinary experts across five dimensions: scientific rationale, personalization, comprehensiveness, safety, and feasibility. Each dimension will be rated on a 5-point Likert scale from 1 to 5. The five dimension scores will be summed to generate an overall quality score ranging from 5 to 25, with higher scores indicating better overall prescription quality. For each prescription, the mean overall score across the six experts will be used for analysis.
Subgroup-Specific Scientific Rationale and Safety Scores of Model-Generated Exercise Prescriptions
Model performance will be evaluated across predefined subgroups based on age, breast cancer stage, molecular subtype, surgical procedure, and treatment modality. Scientific rationale and safety will each be rated on a 1-5 Likert scale, with higher scores indicating better performance. Differences across models and subgroups will be assessed using two-way ANOVA or generalized linear models, including interaction terms between model type and patient characteristics.
Click on the lead sponsor to view all of their trials.
This trial is not recruiting at the moment. You can still explore other options:
