Deep Learning-Based Radiomics Models for Early Recurrence Prediction of Hepatocellular Carcinoma with Multi-phase CT Images and Clinical Data.

Researchers

Journal

Modalities

Models

Abstract

Hepatocellular carcinoma (HCC) is the fifth most common malignancy in the world and the second most common cause of cancer-related death. By surgically removing hepatocellular carcinoma, the patients may have the early recurrence within one year. Recently, CT radiomics signatures have been demonstrated to be effective on predicting early recurrence of HCC. However, the radiomics signatures are based on hand-crafted low-level features, such as density and texture. In this paper, we propose a deep learning-based radiomics approach for predicting early recurrence of hepatocellular carcinoma with multi-phase computed tomography (CT) images. We also propose several models to combine the high-level radiomics features with clinical data to improve the prediction accuracy. The area under the curve (AUC) of receiver operating characteristic curve (ROC) is 0.825.