Can Machine Learning Predict Favorable Outcome After Radiofrequency Ablation of Hepatocellular Carcinoma?

Abstract

Purpose: The standard practice for limited-stage hepatocellular carcinoma (HCC) is the resection or the use of local ablative techniques, such as radiofrequency ablation (RFA). The outcome after RFA depends on a complex interaction between the patient's general condition, hepatic function, and disease stage. In this study, we aimed to explore using a machine learning model to predict the response.

Patients and methods: A retrospective study was conducted for patients with RFA for a localized HCC between 2018 and 2022. The collected clinical, radiologic, and laboratory data were explored using Python and XGBoost. They were split into a training set (70%) and a validation set (30%). The primary end point of this study was to predict the probability of achieving favorable outcomes 12 months after RFA. Favorable outcomes were defined as the patient was alive and HCC was controlled.

Results: One hundred and eleven patients were eligible for the study. Males were 78 (70.3%) with a median age of 57 (range of 43-81) years. Favorable outcome was seen in 62 (55.9%) of the patients. The 1-year survival rate and control rate were 94.6%, and 61.3%, respectively. The final model harbored an accuracy and an AUC of 90.6% and 0.95, respectively, for the training set, while they were 78.9% and 0.80, respectively, for the validation set.

Conclusion: Machine learning can be a predictive tool for the outcome after RFA in patients with HCC. Further validation by a larger study is necessary.