Predicting survival after transarterial chemoembolization for hepatocellular carcinoma using a neural network: A...

Area of Science:

• Hepatobiliary Medicine
• Oncology
• Medical Imaging & Technology

Background:

• Transarterial chemoembolization (TACE) is a standard treatment for hepatocellular carcinoma (HCC).
• Predicting patient survival after TACE is challenging, impacting treatment decisions.
• Existing scoring systems (ART, ABCR, SNACOR) have limitations in accuracy.

Purpose of the Study:

• To develop a novel survival prediction model for HCC patients undergoing TACE.
• To utilize machine learning algorithms, specifically artificial neural networks (ANNs).
• To compare the performance of the ANN model against conventional prediction scores.

Main Methods:

• Retrospective analysis of 282 HCC patients treated with TACE (2005-2017).
• Development of an ANN incorporating standard risk score parameters and additional clinical data.
• Training and validation using an 80:20 data split.

Main Results:

• The ANN demonstrated strong 1-year survival prediction (AUC 0.77 ± 0.13, internal validation AUC 0.83 ± 0.06).
• ANN outperformed SNACOR (0.73), ABCR (0.70), and ART (0.54) in head-to-head comparison.
• The difference was statistically significant compared to ART (P < .001).

Conclusions:

• Artificial neural networks offer superior survival prediction for HCC patients post-TACE compared to traditional scores.
• ANN-based models can aid clinicians in determining optimal treatment strategies.
• Integration of ANNs into clinical practice could enhance patient management and outcomes.