JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Predicting perineural invasion of intrahepatic cholangiocarcinoma based on CT: a multicenter study

  • 0Department of Radiology, First Affiliated Hospital of Sun Yat-sen University, Guangzhou, China.
Abdominal Radiology (new York) +

|

September 1, 2025

|

September 1, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

Machine learning models accurately predict perineural invasion (PNI) in intrahepatic cholangiocarcinoma (ICC) using CT scans. This aids personalized treatment decisions for patients with this liver cancer.

Area Of Science

  • Hepatobiliary surgery
  • Medical imaging
  • Machine learning in oncology

Background

  • Perineural invasion (PNI) is a critical prognostic factor in intrahepatic cholangiocarcinoma (ICC).
  • Accurate preoperative prediction of PNI is challenging but essential for guiding treatment strategies.
  • Current diagnostic methods for PNI often lack sufficient preoperative accuracy.

Purpose Of The Study

  • To assess the feasibility of using machine learning models for preoperative prediction of PNI in ICC.
  • To identify key clinical and CT image features predictive of PNI in ICC.
  • To develop and validate AI-driven models for improved PNI prediction in ICC.

Main Methods

  • A multi-institutional study involving 199 patients with histologically confirmed ICC.
  • Clinical and CT image features were analyzed using LASSO for feature selection.
  • Machine learning models (MLP, RF, SVM, LR, XGBoost) were constructed and validated internally, externally, and prospectively.

Main Results

  • Tumor location, bile duct dilatation, and arterial enhancement pattern were significant predictors of PNI.
  • Machine learning models demonstrated high predictive performance with AUCs exceeding 0.86 in the training cohort.
  • Model performance remained robust across internal, external, and prospective validation cohorts, with AUCs ranging from 0.72 to 0.84.

Conclusions

  • Machine learning models integrating CT features can accurately predict PNI in ICC preoperatively.
  • These AI-based predictive tools can support individualized clinical decision-making for ICC patients.
  • The findings suggest a potential role for ML in refining treatment strategies for ICC.

Keywords: