Download PDF
Open Access
CC BY-NC-ND 4.0 · Indian J Radiol Imaging
DOI: 10.1055/s-0045-1809372
Original Article

Predicting Response to Transarterial Chemoembolization in Hepatocellular Carcinoma Using Machine Learning Models

Authors

  • Niharika Dutta

    1   Department of Radiodiagnosis and Imaging, Postgraduate Institute of Medical Education and Research, Chandigarh, India

  • Pankaj Gupta 

    1   Department of Radiodiagnosis and Imaging, Postgraduate Institute of Medical Education and Research, Chandigarh, India

Funding None.
Further Information
Also available at
  Permissions and Reprints
Preview

Abstract

Background

Hepatocellular carcinoma (HCC) is a leading cause of cancer-related deaths, with transarterial chemoembolization (TACE) being a key treatment for intermediate-stage cases. Accurate prediction of TACE response remains challenging, prompting the exploration of machine learning (ML) models.

Aim

This study aims to investigate ML models for predicting the response to TACE in HCC patients.

Materials and Methods

We utilized the public “WAW-TACE” data set. This data set comprises clinical data and multiphasic computed tomography (CT) images with their corresponding masks. We divided this data set randomly into 183 training and validation cases and 50 held-out test cases. Four models were trained: (A) clinical model incorporating demographic and laboratory parameters, (B) radiomic model using PyRadiomics-extracted features, (C) deep neural network (DNN) using multiphasic CT images processed with MaskedAttentionViT, and (D) combined clinicoradiological model. Performance was assessed using fivefold cross-validation and testing on a held-out data set to predict a lack of response to TACE.

Results

There were 64 (37%) responders and 109 (63%) nonresponders in the training set. There were 13 (26%) responders and 37 (74%) nonresponders in the test set. In the held-out test set, the clinical support vector machine model achieved an accuracy of 70%, sensitivity of 78.9%, specificity of 50%, and area under the curve (AUC) of 0.778 for predicting failure of TACE. The radiomic logistic regression model demonstrated an accuracy of 76.1%, sensitivity of 85.4%, specificity of 18.2%, and AUC of 0.740. The DNN had an accuracy of 63%, sensitivity of 65.7%, specificity of 54.5%, and AUC of 0.601. The combined clinicoradiological model yielded an accuracy of 55.6%, sensitivity of 50%, specificity of 72.7%, and AUC of 0.639.

Conclusion

We utilized a multimodal approach to predict response to TACE in HCC patients. Further optimization and multicenter data sets are required to enhance predictive accuracy further.

Keywords

computed tomography - deep neural network - hepatocellular carcinoma - machine learning - transarterial chemoembolization - treatment response

Authors' Contributions

N.D.: Methodology, writing - original draft, and writing - review and editing.

P.G.: Conceptualization, methodology, writing - original draft, writing - review and editing, and formal analysis.




Publication History

Article published online:
27 May 2025

© 2025. Indian Radiological Association. This is an open access article published by Thieme under the terms of the Creative Commons Attribution-NonDerivative-NonCommercial License, permitting copying and reproduction so long as the original work is given appropriate credit. Contents may not be used for commercial purposes, or adapted, remixed, transformed or built upon. (https://creativecommons.org/licenses/by-nc-nd/4.0/)

Thieme Medical and Scientific Publishers Pvt. Ltd.
A-12, 2nd Floor, Sector 2, Noida-201301 UP, India