Endoscopic ultrasound-based radiomics nomogram for preoperative predicting patients with early esophageal squamous cell carcinoma：a multi-center study
10.3760/cma.j.cn131148-20240723-00403
- VernacularTitle:基于内镜超声的影像组学列线图术前预测早期食管鳞状细胞癌的多中心研究
- Author:
Yajing CHEN
1
;
Shuhan SUN
;
Shumei MIAO
;
Xiaoyan HE
;
Xiaoying ZHOU
;
Feihong YU
Author Information
1. 南京医科大学第一附属医院超声科，南京210029
- Publication Type:Journal Article
- Keywords:
Endoscopic ultrasound;
Radiomics;
Esophageal squamous cell carcinoma;
Nomogram
- From:
Chinese Journal of Ultrasonography
2025;34(1):56-64
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To assess the predictive performance of a nomogram model integrating endoscopic ultrasound（EUS）radiomic features with clinical variables for distinguishing early esophageal squamous cell carcinoma（ESCC）from non-cancerous lesions.Methods:Clinical and imaging data from 454 patients who underwent EUS for suspected esophageal malignancies were retrospectively collected in the First Affiliated Hospital of Nanjing Medical University（training cohort， n = 323）and Dongyang People's Hospital（external validation cohort， n = 131）from January 2020 to November 2023. Independent clinical predictors of early ESCC were identified using univariable and multivariable Logistic regression analyses to establish a clinical model. Pearson correlation and Least Absolute Shrinkage and Selection Operator（LASSO）algorithms were used to construct a radiomics model. A combined model integrating radiomics scores and clinical predictors was developed and visualized as a nomogram. The predictive performance of each model was assessed using the area under the ROC curve（AUC），and calibration curves were used to evaluate the model's fitting capability. Results:The training set and validation set indicated that there were statistically significant differences in age，smoking history and lesion location between the early ESCC group and the non-cancerous lesion change group（all P < 0.05）. According to univariate and multivariate Logistic regression analysis，age（ OR = 1.039，95% CI = 1.003–1.077， P = 0.036）and smoking（ OR = 2.358，95% CI = 1.270 - 4.376， P = 0.007）were identified as independent predictors and used to develop the clinical model，with AUCs of 0.608 and 0.694 in the training and validation cohorts，respectively. Fourteen optimal radiomic features were selected to construct the radiomics model，with AUCs of 0.881 and 0.807 in the training and validation cohorts，respectively. The combined nomogram model demonstrated superior predictive performance with AUCs of 0.893 and 0.830，sensitivities of 82.5% and 79.1%，and specificities of 82.2% and 81.3% in the training and validation cohorts，respectively. Conclusions:The EUS-based nomogram model demonstrates optimal predictive performance and can serve as a non-invasive tool to assist endoscopists in distinguishing early ESCC from non-cancerous lesions.
