4:03 PM Abstract No. 305 Progression-free survival and overall survival after transcatheter arterial chemoembolization in Barcelona Clinic Liver Cancer stage B hepatocellular carcinoma patients: a prediction model approach by recurrent neural network

Abstract

A database for BCLC B hepatocellular carcinoma (HCC) patients who had received transcatheter arterial chemoembolization (TACE) was used create a prediction models for 1-, 2-, 3-, 4-, and 5-year progression-free survival (PFS) and overall survival (OS) based on a clinical parameters for the patient group. Recurrent neural network (RNN) was used to predict a year based OS and PFS. The thirty-four clinical features and imaging evaluation with modified RECIST as primary response after a first TACE were used with an RNN model. Data for 87 HCC patients with histories of 1-, 2-, 3-, 4- and 5- year PFS and OS after TACE respectively, were chosen from the HCC patient database. RNN was used to make the recurrent network for survival, and patients were randomly sorted into a training set (70%) and a test set (30%). Receiver operating characteristic (ROC) curves and area under the curves (AUCs) were evaluated, and we compared the prediction of each year survival in both OS and PFS. At 1-, 2-, 3-, 4- and 5-year OS, ROC at each year was as follows: 0.48 at 1 –year OS, 0.67 at 2-year OS, 0.83 at 3-year OS, 0.98 at 4-year OS, 0.86 at 5-year OS. At 1-, 2-, 3-, 4- and 5-year PFS, 0.97 at 1-year PFS, 1.0 at 2-year PFS, 1.0 at 3-year PFS, 1.0 at 4-year PFS, 0.96 at 5-year PFS. The RNN for survival model outperformed at PFS compared with OS in terms of prediction accuracy. This study shows the feasibility of using RNN decision making support systems for predicting the PFS and OS after TACE for patients with BCLC B HCC.