文献类型：期刊文献

英文题名：Preoperative prediction of tumor budding grade in rectal cancer by combining APT histogram analysis and ADC MRI

作者：Zhang, Yingying[1];Du, Yunxia[2];Huang, Jinghuan[1];Yang, Ou[1];Gong, Juntao[1];Zhang, Xiaoyue[3];Sun, Yun[1];Li, Feixiang[1];Wang, Jiaqi[1];Huang, Gang[4]

第一作者：张延英;张彦彦;张芸燕

通信作者：Huang, G[1]

机构：[1]Gansu Univ Chinese Med, Lanzhou, Peoples R China;[2]Univ Chinese Med, Lanzhou Petrochem Gen Hosp, Affiliated Hosp Gansu 4, Lanzhou, Peoples R China;[3]Philips Healthcare, Dept Clin & Tech Support, Xian, Peoples R China;[4]Gansu Prov Hosp, Dept Radiol, Lanzhou, Peoples R China

第一机构：甘肃中医药大学

通信机构：[1]corresponding author), Gansu Prov Hosp, Dept Radiol, Lanzhou, Peoples R China.

年份：2026

卷号：10

期号：1

外文期刊名：EUROPEAN RADIOLOGY EXPERIMENTAL

收录：Scopus(收录号：2-s2.0-105039676844);WOS:【ESCI(收录号:WOS:001771673400004)】；

基金：This study was supported by the Provincial Nature Science Foundation of Gansu (grant no. 24JRRA1054) and the Gansu Provincial Hospital Intramural Research Fund (grant no. 23GSSYD-6).

语种：英文

外文关键词：Biomarkers; Diffusion magnetic resonance imaging; Machine learning; Neoplasm staging; Rectal neoplasms

摘要：Objective Tumor budding (TB) is a histopathological marker of aggressive behavior and poor prognosis in rectal cancer (RC), yet not reliably evaluated preoperatively. We assessed whether histogram features from amide proton transfer-weighted (APTw) imaging and apparent diffusion coefficient (ADC) maps could serve as noninvasive biomarkers for preoperative TB grade prediction. Materials and methods This retrospective study included 204 patients with RC from June 2023 to May 2025, divided into a training cohort (n = 133) and a validation cohort (n = 71) using a temporal split. All patients underwent preoperative APTw and diffusion-weighted imaging, and TB grade was determined histopathologically. Histogram features were extracted from whole-tumor volumes on APTw and ADC maps. Feature selection used a machine learning-based classifier, followed by univariate and multivariate logistic regression to identify independent predictors. SHapley Additive exPlanations (SHAP) were applied for interpretability, and a nomogram integrating histogram and clinical variables was constructed. Results Five key histogram features (ADC-90%, ADC-Minimum, ADC-Range, APTw-10%, and APTw-Median) were selected. The histogram model achieved areas under the curve (AUROCs) of 0.85 (95% confidence interval [CI]: 0.79-0.92) and 0.86 (95% CI: 0.78-0.95) in the training and validation cohorts. SHAP analysis identified ADC-90% and ADC-Minimum as the most influential predictors. The combined model with histogram and clinical factors showed improved performance, with AUROCs of 0.88 (95% CI: 0.82-0.94) and 0.87 (95% CI: 0.79-0.96). Conclusion APTw and ADC histogram features can independently predict TB grade in RC. The combined model, integrating both histogram and clinical features, further enhanced preoperative predictive accuracy.