Predicting hepatocellular carcinoma response to TACE: A machine learning study based on 2.5D CT imaging and deep...

Area of Science:

• Radiology
• Oncology
• Artificial Intelligence in Medicine

Background:

• Accurate prognosis prediction for hepatocellular carcinoma (HCC) patients undergoing transarterial chemoembolization (TACE) is crucial for treatment planning.
• Current methods lack objectivity, necessitating advanced predictive tools.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting TACE response in HCC patients.
• To utilize deep features extracted from 2.5D CT images for prognosis prediction.

Main Methods:

• A public dataset (TCIA) was used for ResNet50 transfer learning and ML model construction.
• An external testing dataset of 26 patients treated at the institution was used for validation.
• 2.5D images were constructed from axial, sagittal, and coronal CT views; ResNet50 served as a feature extractor.

Main Results:

• Machine learning models (RFC, SVC, LR, XGB) demonstrated strong performance on the external testing dataset.
• Area Under the Curve (AUC) values ranged from 0.89 to 0.91.
• Accuracy ranged from 0.79 to 0.81, and F1-scores ranged from 0.75 to 0.79.

Conclusions:

• The developed ML model, leveraging deep features from 2.5D CT images, shows significant potential for predicting HCC patient prognosis after TACE.
• This approach can aid clinicians in making more informed treatment decisions for HCC patients.