Scholarlink

Background <p>This study focuses on extracting preoperative vertebral quantitative CT (QCT) radiomic features and developing models to predict changes in bone mineral density (BMD) after sleeve gastrectomy (SG).</p> Methods <p>A retrospective analysis was conducted on 203 patients who underwent SG at Qujing Second People's Hospital between June 2022 and February 2024. Patients were divided into two groups based on changes in lumbar vertebra 1 QCT values: BMD decreased and BMD stable/increased. Data were randomly split into training and test sets (7:3 ratio) using stratified sampling. Radiomic features were extracted and normalized, and feature selection was performed using ICC, variance thresholding, mutual information, and LASSO. XGBoost models were built for clinical, radiomic, and combined data, with performance evaluated using ROC curves, AUC, and decision curve analysis (DCA).</p> Results <p>Significant differences were observed in BMI, erector spinae average CT value, AST, and ALT between groups. Based on clinical and radiomic features, the AUC values of the XGBoost models in the training and test sets were as follows: clinical model 0.94, 0.88; radiomic model 0.98, 0.96; combined model 0.97, 0.96. DCA showed that the combined model provided the highest net benefit across all threshold values.</p> Conclusion <p>Vertebral QCT combined with clinical features can effectively predict postoperative BMD changes after SG.</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

Quantitative CT Imaging Radiomics-Based Prediction of Bone Mineral Density Changes After Sleeve Gastrectomy

  • 期刊论文

  • 发表日期：

    2025年6月19日

查看全文
  • Deyao Hu,
  • Jilu Ruan,
  • Chengjian Liu,
  • Zhengrong Liang,
  • Xuetao Mu

摘要

Background

This study focuses on extracting preoperative vertebral quantitative CT (QCT) radiomic features and developing models to predict changes in bone mineral density (BMD) after sleeve gastrectomy (SG).

Methods

A retrospective analysis was conducted on 203 patients who underwent SG at Qujing Second People's Hospital between June 2022 and February 2024. Patients were divided into two groups based on changes in lumbar vertebra 1 QCT values: BMD decreased and BMD stable/increased. Data were randomly split into training and test sets (7:3 ratio) using stratified sampling. Radiomic features were extracted and normalized, and feature selection was performed using ICC, variance thresholding, mutual information, and LASSO. XGBoost models were built for clinical, radiomic, and combined data, with performance evaluated using ROC curves, AUC, and decision curve analysis (DCA).

Results

Significant differences were observed in BMI, erector spinae average CT value, AST, and ALT between groups. Based on clinical and radiomic features, the AUC values of the XGBoost models in the training and test sets were as follows: clinical model 0.94, 0.88; radiomic model 0.98, 0.96; combined model 0.97, 0.96. DCA showed that the combined model provided the highest net benefit across all threshold values.

Conclusion

Vertebral QCT combined with clinical features can effectively predict postoperative BMD changes after SG.