Radiomics and deep learning for predicting short-term outcomes of neoadjuvant therapy in esophageal cancer
10.3760/cma.j.cn113030-20250304-00080
- VernacularTitle:影像组学结合深度学习预测食管癌新辅助治疗近期疗效
- Author:
Nana YU
1
;
Linrui LI
;
Mengyu HAN
;
Xiaoyang LI
;
Liting QIAN
Author Information
1. 山东大学齐鲁医学院安徽省立医院肿瘤放射治疗科，合肥 230001
- Publication Type:Journal Article
- Keywords:
Esophageal neoplasms;
Deep learning;
Radiomics;
Neoadjuvant chemoradiotherapy;
Pathological complete response
- From:
Chinese Journal of Radiation Oncology
2025;34(12):1199-1207
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To explore the predictive value of models based on clinical parameters, deep learning radiomics (DLR) from CT images, and traditional handcrafted radiomics (HCR) in assessing pathological complete response (pCR) after neoadjuvant radiotherapy combined with medical therapy in patients with esophageal cancer.Methods:A retrospective study was conducted on 130 patients with locally advanced esophageal cancer who underwent neoadjuvant radiotherapy combined with medical therapy followed by surgery at the First Affiliated Hospital of the University of Science and Technology of China from August 1, 2018, to August 31, 2024. Patients were randomly divided into a training set ( n=91) and a validation set ( n=39) at a ratio of 7:3. Logistic regression analysis was performed to identify clinical independent risk factors associated with pCR. DLR and HCR features were extracted from the tumor and the 5 mm peritumoral region on planning CT images. Features for modeling were selected using t-test, Mann-Whitney U test or Fisher exact probability method, least absolute shrinkage and selection operator (LASSO) regression to calculate the radiomics score (Rad-score). A nomogram was then constructed by integrating the clinical risk factors. The predictive performance of each model was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, and decision curve analysis (DCA) to assess clinical benefits. Results:Multivariate logistic regression analysis identified body weight ( OR=1.101, 95% CI: 1.029-1.177, P=0.005) and lymph node positivity ( OR=0.100, 95% CI: 0.014-0.727, P=0.023) as independent predictors of pCR. The peritumoral DLR-HCR model showed superior predictive performance, with AUCs of 0.870 (95% CI: 0.799-0.942) in the training set and 0.866 (95% CI: 0.750-0.982) in the validation set. The combined model incorporating clinical parameters achieved the best performance, with AUCs of 0.903 (95% CI: 0.845-0.962) and 0.888 (95% CI: 0.782-0.994) in the training and validation sets, respectively. Conclusions:The combined model integrating peritumoral DLR-HCR features with clinical parameters provides excellent predictive value for pCR after neoadjuvant radiotherapy combined with medical therapy in esophageal cancer and offers valuable guidance for personalized treatment strategies.
