Scholarlink

Purpose <p>Muscle loss after radiotherapy is associated with poor overall survival in patients with oral cavity cancer. In this study, we aimed to develop a machine learning model for predicting muscle loss after radiotherapy.</p> Methods <p>This study included patients with oral cavity cancer who underwent surgery and post-operative radiotherapy at two tertiary centers between 2010 and 2020. Muscle loss was determined by comparing pre- and post-radiotherapy computed tomography scans. The Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Categorical Boosting (CatBoost) models were trained to predict muscle loss using clinical and toxicity features. Model performance was evaluated using the area under the curve (AUC). The SHapley Additive exPlanations (SHAP) method was used to interpret the model.</p> Results <p>Of 903 eligible patients (median age: 55 years), 572 and 331 were in the derivation and external validation cohorts, with 144 (25.2%) and 81 (24.5%) patients experiencing muscle loss after radiotherapy, respectively. The CatBoost model achieved the highest AUC compared to the RF and XGBoost models in the external validation cohort (AUC: 0.993, 0.991, and 0.989, respectively). Mini Nutritional Assessment score, anorexia, and dysphagia were identified as the top three contributing features for muscle loss prediction. The SHAP force plot provided a personalized model prediction interpretation for each patient.</p> Conclusions <p>A machine learning model based on clinical and toxicity data could predict muscle loss after radiotherapy for oral cavity cancer. Personalized predictions could assist clinicians in guiding interventions for muscle mass preservation.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning for predicting muscle loss after radiotherapy using clinical and toxicity data in oral cavity cancer

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Ning-Hsiang Weng,
  • Jhen-Bin Lin,
  • Ya-Ting Jan,
  • Yi-Hsuan Lin,
  • Yi-Shing Leu,
  • Yu-Jen Chen,
  • Kun-Pin Wu,
  • Jie Lee

摘要

Purpose

Muscle loss after radiotherapy is associated with poor overall survival in patients with oral cavity cancer. In this study, we aimed to develop a machine learning model for predicting muscle loss after radiotherapy.

Methods

This study included patients with oral cavity cancer who underwent surgery and post-operative radiotherapy at two tertiary centers between 2010 and 2020. Muscle loss was determined by comparing pre- and post-radiotherapy computed tomography scans. The Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Categorical Boosting (CatBoost) models were trained to predict muscle loss using clinical and toxicity features. Model performance was evaluated using the area under the curve (AUC). The SHapley Additive exPlanations (SHAP) method was used to interpret the model.

Results

Of 903 eligible patients (median age: 55 years), 572 and 331 were in the derivation and external validation cohorts, with 144 (25.2%) and 81 (24.5%) patients experiencing muscle loss after radiotherapy, respectively. The CatBoost model achieved the highest AUC compared to the RF and XGBoost models in the external validation cohort (AUC: 0.993, 0.991, and 0.989, respectively). Mini Nutritional Assessment score, anorexia, and dysphagia were identified as the top three contributing features for muscle loss prediction. The SHAP force plot provided a personalized model prediction interpretation for each patient.

Conclusions

A machine learning model based on clinical and toxicity data could predict muscle loss after radiotherapy for oral cavity cancer. Personalized predictions could assist clinicians in guiding interventions for muscle mass preservation.