Scholarlink

Purpose <p>To development of a preoperative CT-based radiomics model for predicting muscle invasion in patients with upper tract urothelial carcinoma below T3 stage.</p> Methods <p>163 consecutive patients who underwent radical nephroureterectomy for stage pT1–2 UTUC were retrospectively enrolled two medical centers (116 patients in training data and 47 patients in external validation data). Lesion segmentation, extraction and selection of radiomic features on pre-surgical CT urography, development and validation of predictive models were performed. Risk stratification of UTUC was evaluated. The diagnostic performance of the radiomics model and risk stratification was analyzed. Reference standard was histopathological analysis.</p> Results <p>Among 163 patients (mean age, 52 years ± 9 [standard deviation], 97 men), 61.5% had pT2 grade tumors. 1165 features with intraclass coefficients &gt; 0.75 were retained for least absolute shrinkage and selection operator (LASSO) regression. Nine radiomic features with non-zero coefficients on LASSO regression were selected from the training dataset and used for constructing the radiomics model. Good discrimination capability of the predictive model was observed, as AUCs were 0.859 (95% CI, 0.782–0.917) in the training dataset and 0.821 (95% CI, 0.682–0.918) in the validation dataset, respectively. Based on judgement by the model, When the tumor length diameter &gt; 3&#xa0;cm, combining ureteroscopy biopsy would improve sensitivity and NPV to 0.86 (95% CI, 0.776–0.922), 0.81 (95% CI, 0.714–0.903).</p> Conclusion <p>The preoperative radiomics model showed promising diagnostic performance in predicting UTUC muscle invasion. This could help patients receive more accurate risk classification, especially help patients avoiding radical nephroureterectomy.</p> Graphical abstract <p></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

Preoperative CT-based radiomics model for predicting muscle invasion in patients with upper tract urothelial carcinoma below T3 stage

  • 期刊论文

  • 发表日期：

    2025年5月17日

查看全文
  • Han-Mei Zhang,
  • Yi Wang,
  • Zi-Xing Huang,
  • Yu-Xi Liu,
  • Li Liu,
  • Yi-Ge Bao,
  • Xiang Cai,
  • Tao Wu,
  • Qian Xu,
  • Xiang-Lan Zhu,
  • Hong-Kun Yin,
  • Hui-Ling Zhang,
  • Fang Yuan,
  • Bin Song

摘要

Purpose

To development of a preoperative CT-based radiomics model for predicting muscle invasion in patients with upper tract urothelial carcinoma below T3 stage.

Methods

163 consecutive patients who underwent radical nephroureterectomy for stage pT1–2 UTUC were retrospectively enrolled two medical centers (116 patients in training data and 47 patients in external validation data). Lesion segmentation, extraction and selection of radiomic features on pre-surgical CT urography, development and validation of predictive models were performed. Risk stratification of UTUC was evaluated. The diagnostic performance of the radiomics model and risk stratification was analyzed. Reference standard was histopathological analysis.

Results

Among 163 patients (mean age, 52 years ± 9 [standard deviation], 97 men), 61.5% had pT2 grade tumors. 1165 features with intraclass coefficients > 0.75 were retained for least absolute shrinkage and selection operator (LASSO) regression. Nine radiomic features with non-zero coefficients on LASSO regression were selected from the training dataset and used for constructing the radiomics model. Good discrimination capability of the predictive model was observed, as AUCs were 0.859 (95% CI, 0.782–0.917) in the training dataset and 0.821 (95% CI, 0.682–0.918) in the validation dataset, respectively. Based on judgement by the model, When the tumor length diameter > 3 cm, combining ureteroscopy biopsy would improve sensitivity and NPV to 0.86 (95% CI, 0.776–0.922), 0.81 (95% CI, 0.714–0.903).

Conclusion

The preoperative radiomics model showed promising diagnostic performance in predicting UTUC muscle invasion. This could help patients receive more accurate risk classification, especially help patients avoiding radical nephroureterectomy.

Graphical abstract