Scholarlink

Objective <p>To assess the value of multiparametric MRI (mpMRI)-derived radiomic signatures and a combined model for non-invasive prediction of human epidermal growth factor receptor 2 (HER2) expression in bladder cancer (BCa).</p> Methods <p>A total of 113 BCa patients with preoperative pelvic mpMRI were retrospectively enrolled. Radiomic features were extracted from T2WI, DWI, DCE and their combinations. Five machine learning algorithms were used to construct radiomic models. A combined model and a nomogram were developed by integrating radiomic signatures and clinicoradiological variables.</p> Results <p>The T2WI + DWI + DCE-based RandomForest model achieved the best performance, with an AUC of 0.877 in the training cohort and 0.754 in the validation cohort. Age, risk group, and maximum tumor diameter were independent predictors of HER2 overexpression. The combined model yielded AUCs of 0.808 and 0.870 in the training and validation cohorts, respectively.</p> Conclusion <p>mpMRI radiomics can non-invasively predict HER2 expression in BCa. The combined nomogram shows good clinical utility, supporting personalized treatment planning for BCa patients.</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

Multiparametric MRI-derived radiomic signatures enable noninvasive prediction of HER2 expression in bladder cancer

  • 期刊论文

  • 发表日期：

    2026年7月3日

查看全文
  • Yali Wang,
  • Tonglei Zhao,
  • Baotai Liang,
  • Lan Zhen,
  • Yiming Ding,
  • Yanji Jiang,
  • Jianping Wu,
  • Yuan Meng,
  • Weipu Mao,
  • Ming Chen

摘要

Objective

To assess the value of multiparametric MRI (mpMRI)-derived radiomic signatures and a combined model for non-invasive prediction of human epidermal growth factor receptor 2 (HER2) expression in bladder cancer (BCa).

Methods

A total of 113 BCa patients with preoperative pelvic mpMRI were retrospectively enrolled. Radiomic features were extracted from T2WI, DWI, DCE and their combinations. Five machine learning algorithms were used to construct radiomic models. A combined model and a nomogram were developed by integrating radiomic signatures and clinicoradiological variables.

Results

The T2WI + DWI + DCE-based RandomForest model achieved the best performance, with an AUC of 0.877 in the training cohort and 0.754 in the validation cohort. Age, risk group, and maximum tumor diameter were independent predictors of HER2 overexpression. The combined model yielded AUCs of 0.808 and 0.870 in the training and validation cohorts, respectively.

Conclusion

mpMRI radiomics can non-invasively predict HER2 expression in BCa. The combined nomogram shows good clinical utility, supporting personalized treatment planning for BCa patients.