Contrast-enhanced computed tomography radiomics for the preoperative prediction model of microvascular invasion in intrahepatic cholangiocarcinoma
10.3760/cma.j.cn115396-20240701-00205
- VernacularTitle:增强CT影像组学用于肝内胆管癌微血管侵犯的术前预测模型
- Author:
Zheyu ZHOU
1
;
Shuya CAO
;
Chunlong ZHAO
;
Qiaoyu LIU
;
Xiaoliang XU
;
Chaobo CHEN
Author Information
1. 中国医学科学院北京协和医学院，北京 100730
- Keywords:
Bile duct neoplasms;
Forecasting;
Models, statistical;
Intrahepatic cholangiocarcinoma;
Microvascular invasion;
Radiomics;
Contrast-enhanced computed tomogr
- From:
International Journal of Surgery
2024;51(8):511-516
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To predict the status of microvascular invasion (MVI) in intrahepatic cholangiocarcinoma (ICC) patients preoperatively based on the radiomics analysis of contrast-enhanced CT to provide imaging evidence for early identification of patients at high risk of recurrence.Methods:Clinical data of 40 ICC patients who underwent radical hepatectomy at Nanjing Drum Tower Hospital from January 2021 to May 2024 were retrospectively collected. Patients were divided into the MVI group ( n=8) and the non-MVI group ( n=32) according to the MVI status of the postoperative pathology report. Whether there were differences in each pathological index between the groups and the efficacy of radiomics analysis of contrast-enhanced CT for the preoperative prediction of MVI were analyzed. The regions of interest (ROI) were outlined on the arterial and venous phase images using the 3D Slicer software. Then, radiomics features were extracted from each ROI based on Python. Finally, the LASSO regression and glm function were used to screen radiomics features and establish a prediction model based on the R language. The established predictive model′s diagnostic efficacy, calibration, and net clinical benefit were evaluated using the receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA), respectively. Normally distributed measurement data were expressed as mean±standard deviation ( ± s) and compared using the t-test. Count data were expressed as frequency and compared using the chi-square test. Results:Patients in the MVI group had more poorly differentiated tumors and a significantly higher proportion of lymph node metastases ( P<0.05). The established radiomics prediction model included six features, 1 first-order statistical feature and 5 gray texture features. The area under the ROC curve was 0.87, the sensitivity was 75.0%, and the specificity was 90.6%. The calibration curve showed good agreement between the predicted MVI and actual MVI status, and the decision curve demonstrated that the model could provide a large net clinical benefit. Conclusion:Radiomics analysis of contrast-enhanced CT can identify the MVI status of ICC patients preoperatively and aid in clinical decision-making, providing vital evidence for individualized and precise treatment of ICC.
