Noninvasive Imaging-based Machine Learning Algorithm To Identify Progressive Disease In Advanced Hepatocellular Carcinoma Receiving Second-line Systemic Therapy

This study examined the use of radiomic features extracted from initial CT images before TKI-PD-1 treatment to predict the response of hepatocellular carcinoma (HCC) to treatment. A radiomics-based support vector machine (SVM) machine learning algorithm using ten LASSO filtered features was used to predict the treatment response with an accuracy of 81.8%, sensitivity of 100.0%, specificity of 43.7%, precision of 78.8%, and F1 score of 88.0% in the training cohort and 69.1%, 95.0%, 20.0%, 70.6%, and 80.0%, respectively, in the testing cohort. Additionally, four out of ten features were found to be significantly associated with overall survival. Cox regression analysis incorporating radiomic features and clinical variables identified seven factors associated with survival, two of which were favorable survival factors.