CT radiomics combined with CT and preoperative pathological features for predicting postoperative early recurrence of local advanced esophageal squamous cell carcinoma
10.13929/j.issn.1003-3289.2024.06.014
- VernacularTitle:术前CT影像组学联合CT及病理特征预测局部进展期食管鳞癌术后早期复发
- Author:
Jingjing XING
1
;
Yiyang LIU
;
Yue ZHOU
;
Pengchao ZHAN
;
Rui WANG
;
Yaru CHAI
;
Peijie LYU
;
Jianbo GAO
Author Information
1. 郑州大学第一附属医院放射科,河南郑州 450052
- Keywords:
esophageal neoplasms;
carcinoma,squamous cell;
X-ray computed,tomography;
radiomics;
neoplasm recurrence,local
- From:
Chinese Journal of Medical Imaging Technology
2024;40(6):863-868
- CountryChina
- Language:Chinese
-
Abstract:
Objective To investigate the value of CT radiomics combined with CT and preoperative pathological features for predicting postoperative early recurrence(ER)of local advanced esophageal squamous cell carcinoma(LAESCC).Methods Data of 334 patients with LAESCC were retrospectively analyzed.The patients were divided into training set(n=234)and verification set(n=100)at the ratio of 7:3 and were followed up to observe ER(recurrence within 12 months after surgery)or not.Univariate and multivariate logistic regression were used to analyze clinical,CT and preoperative pathological features of LAESCC in patients with or without ER in training set.The independent risk factors of ER were screened,and a CT-preoperative pathology model was constructed.Based on venous phase CT in training set,the radiomics features of lesions were extracted and screened to establish radiomics model,and finally a combined model was established based on radiomics model and the independent risk factors.Receiver operating characteristic(ROC)curves were drawn,and the area under the curve(AUC)was calculated to evaluate the diagnostic efficacy of each model.Results Among 334 cases,168 were found with but 166 without ER.In training set,117 cases were found with while the rest 117 without ER,while in verification set,51 were found with but 49 without ER.The length of lesions,cT stage and cN stage shown on CT and tumor differentiation degree displayed with preoperative pathology were all independent risk factors for ER of LAESCC(all P＜0.05).The AUC of CT-preoperative pathology model in training set and validation set was 0.759 and 0.783,respectively.Ten best radiomics features of LAESCC were selected,and AUC of the established radiomics model in training set and validation set was 0.770 and 0.730,respectively.The AUC of combined model in training and validation set was 0.838 and 0.826,respectively.The AUC of CT radiomics combined with CT and preoperative pathological features in training set was higher than that of CT-preoperative pathologymodel and radiomics model(both P＜0.01).Conclusion CT radiomics combined with CT and preoperative pathological features could effectively predict postoperative ER of LAESCC.
