A novel MRI radiomics-based nomogram for preoperative prediction of perineural invasion in intrahepatic cholangiocarcinoma
10.19405/j.cnki.issn1000–1492.2026.04.019
- VernacularTitle:基于MRI影像组学列线图术前预测肝内胆管癌神经侵犯
- Author:
Huize SUI
1
;
Zheyu ZHOU
2
;
Shuya CAO
3
;
Xiaoliang XU
1
;
Guoqiang LI
1
Author Information
1. Dept of Hepatic-Biliary-Pancreatic Surgery，The First Affiliated Hospital of Anhui Medical University，Hefei 230022
2. Dept of General Surgery，Nanjing Drum Tower Hospital，Chinese Academy of Medical Sciences & Peking Union Medical College，Nanjing 210008
3. Dept of General Surgery， Suzhou Hospital of Traditional Chinese Medicine，Nanjing University of Chinese Medicine，Suzhou 215009
- Publication Type:Journal Article
- Keywords:
intrahepatic cholangiocarcinoma;
perineural invasion;
magnetic resonance imaging;
radiomics;
nomogram;
preoperative prediction
- From:
Acta Universitatis Medicinalis Anhui
2026;61(4):736-742
- CountryChina
- Language:Chinese
-
Abstract:
ObjectiveTo evaluate a novel nomogram based on contrast-enhanced MRI radiomics combined with clinical variables for the preoperative prediction of perineural invasion （PNI） in intrahepatic cholangiocarcinoma （ICC）. MethodsThe clinical data of 59 ICC patients were retrospectively collected. According to postoperative pathology reports， the patients were divided into the non-PNI group （n = 33） and the PNI group （n = 26）. Regions of interest （ROI） were delineated from five MRI sequences. Radiomics features were then extracted and filtered to select those with the strongest discriminative power for PNI identification. These selected features were used to construct a radiomics model， which subsequently generated a quantitative radiomics score （radiomics score， Radscore）. Univariate analysis was applied to identify clinical variables associated with PNI， and the glm function was subsequently used to construct clinical and combined models. Finally， the models were evaluated using receiver operating characteristic （ROC） curves， calibration curves， and decision curve analysis （DCA）. The combined model was then visualized as a nomogram. ResultsThe clinical model included age， carbohydrate antigen 19-9 （CA19-9）， red blood cell distribution width， and albumin， whereas the Radscore included five radiomic features. The areas under the ROC curves （AUCs） for the clinical and radiomics models were 0.717 （95%CI： 0.586-0.848） and 0.896 （95%CI： 0.820-0.973）， respectively， whereas the combined model further improved its AUC to 0.917 （95% CI：0.848-0.987）. The calibration curves and DCA showed that the nomogram was well calibrated and provided the greatest net clinical benefit. ConclusionThe novel nomogram may serve as a basis for preoperative prediction of PNI status， thereby assisting clinical decision-making and guiding personalized treatment.
