Introduction:
Intrahepatic cholangiocarcinoma (ICC) accounts for 10-15% in primary liver cancer(1). While hepatectomy is the primary treatment, the 5-year postoperative survival rate is only 25-40%, largely due to factors like microvascular invasion (MVI) (2,3). Unlike macrovascular invasion, which can be detected preoperatively, MVI is only visible microscopically after surgery, making preoperative diagnosis challenging(4). Radiomics, a noninvasive and advanced technique, extracts quantitative imaging features from medical images to enhance diagnostic and prognostic accuracy(5). However, existing studies have not constructed a nomogram that integrates MVI status and patient outcomes in ICC(6,7). Therefore, it is reasonable to investigate the potential of radiomic features extracted from multiparametric MRI for preoperative MVI stratification and survival prediction in ICC patients.
Materials and Methods:
From January 2016 to June 2019, we enrolled 249 patients and divided them into a training cohort (174 patients, 48 with MVI and 126 without MVI) and a validation cohort (75 patients, 20 with MVI and 55 without MVI) in a 7:3 ratio. We also included a time-independent test cohort of 47 ICC patients (16 with MVI and 31 without MVI) from July 2019 to January 2020. MRI images were carried out on a 3.0T uMR 770 scanner (United Imaging Healthcare, Shanghai, China). Our radiomics analysis workflow involved tumor segmentation, feature extraction, feature selection, model construction, and model evaluation. Clinical and imaging features were selected using univariable logistic regression, and multivariable logistic regression identified independent MVI status features. We initially created a radiomics model using decision tree (DT), K-nearest neighbor (KNN) algorithm, logistic regression (LR), random forest (RF), and support vector machine (SVM) classifiers. Receiver operating characteristic curves (ROCs) were plotted, and AUCs quantified the models' predictive efficacy, with corresponding accuracy, sensitivity, and specificity calculations. Calibration curves assessed the agreement between predicted MVI status and actual MVI status using the Hosmer–Lemeshow test. Decision curves were plotted to assess the clinical net benefit at different risk thresholds in the three cohorts. Univariate and multivariate Cox regression identified preoperative independent risk factors for overall survival in histological and predicted MVI groups. Survival probabilities were evaluated and depicted through Kaplan–Meier survival analysis and compared using the log-rank test.
Results:
Multivariate analysis identified tumor size (p=0.001; OR=1.024, 95% CI: 1.009-1.039) and intrahepatic duct dilatation (p<0.001; OR=3.236, 95% CI: 1.723-6.077) as independent MVI predictors. We identified 25 stable radiomics features associated with MVI through LASSO regression from six MR sequences. The logistic regression (LR) classifier was selected for the radiomics model due to its stable predictive efficacy compared to other classifiers, exhibiting desirable performance. The Imaging-Radiomics (IR) model, integrating imaging and radiomics, outperformed other models in all three cohorts, with AUCs of 0.890 in the training cohort, 0.885 in the validation cohort, and 0.815 in the test cohort. Calibration and decision curves validated its performance. Overall survival rates at 1, 3, and 5 years were 85.8%, 64.2%, and 54.9%, respectively. Multivariate Cox regression analysis identified histological MVI status, serum CA19-9 levels, and intrahepatic duct dilatation as independent risk factors for overall survival in the histological MVI group. In the IR model-predicted MVI group, the IR model-predicted MVI status and serum CA19-9 were also identified as independent risk factors for overall survival. The median overall survival for positive and negative histological MVI subgroups was 31.8 months and over 78.1 months, respectively (p<0.001). Interestingly, the median overall survival for the positive and negative IR model-predicted MVI subgroups was 32.8 months and over 78.1 months, respectively (p<0.001).
Discussion:
This study marks the first endeavor to predict MVI status and OS preoperatively using preoperative MR imaging and extracted radiomics features in a noninvasive, personalized approach. The imaging model comprises tumor size and intrahepatic duct dilatation, while the radiomics model includes 25 stable radiomics features from six MR sequences. Both the radiomics and IR models outperform the imaging model, with the radiomics model and IR model performing similarly. However, while the radiomics model exhibits perfect predictive efficacy in the training cohort, its AUC decreases significantly in the validation and test cohorts compared to the IR model. This suggests that the IR model offers better reproducibility in the validation and test cohorts. In conclusion, the IR model, with comparable OS predictive power to histological MVI status, offers a more practical and convenient tool for preoperative MVI status and survival prediction.
Conclusion:
The IR model and the associated nomogram based on IR model-predicted MVI status have the potential to be valuable tools for preoperative MVI status stratification and overall survival prediction in ICC patients.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71(3):209-249.

2. Mazzaferro V, Gorgen A, Roayaie S, Droz Dit Busset M, Sapisochin G. Liver resection and transplantation for intrahepatic cholangiocarcinoma. J Hepatol 2020;72(2):364-377.

3. Hue JJ, Rocha FG, Ammori JB, et al. A comparison of surgical resection and liver transplantation in the treatment of intrahepatic cholangiocarcinoma in the era of modern chemotherapy: An analysis of the National Cancer Database. J Surg Oncol 2021;123(4):949-956.

4. Shao C, Chen J, Chen J, Shi J, Huang L, Qiu Y. Histological classification of microvascular invasion to predict prognosis in intrahepatic cholangiocarcinoma. Int J Clin Exp Pathol 2017;10(7):7674-7681.

5. Wei J, Jiang H, Gu D, et al. Radiomics in liver diseases: Current progress and future opportunities. Liver Int 2020;40(9):2050-2063.

6. Ma X, Liu L, Fang J, et al. MRI features predict microvascular invasion in intrahepatic cholangiocarcinoma. Cancer Imaging 2020;20(1):40.

7. Qian X, Lu X, Ma X, et al. A Multi-Parametric Radiomics Nomogram for Preoperative Prediction of Microvascular Invasion Status in Intrahepatic Cholangiocarcinoma. Front Oncol 2022;12:838701.