Deep Fusion Models of Multi-Phase CT and Selected Clinical Data for Preoperative Prediction of Early Recurrence in Hepatocellular Carcinoma

Abstract

Post-operative early recurrence (ER) of hepatocellular carcinoma (HCC) is one of the leading causes of death. The prediction of the ER of HCC before treatment contributes to guiding treatment and follow-up protocols. In recent years, CT radiomics signatures have been proven effective in several studies in predicting early recurrence of HCC, there are still two major challenges. First, the radiomics features extracted were low or mid-level features, which may not fully characterize HCC heterogeneity. Second, the fusion approach of clinical textual data and image information is in little consensus. In this paper, we proposed a deep-learning based prediction model to extract high-level features from the triple-phase CT images and compare its performance with traditional radiomics model and clinical model. The accuracy and area under the curve (AUC) of receiver operating characteristics of three models was 69.52%/0.723, 67.04%/0.64, 76.03%/0.75, respectively. In addition, we proposed four fusion models to combine clinical data and highlevel features. Among them, Fusion model D performed best, achieving a higher prediction accuracy of 78.66% and AUC of 0.8248. Moreover, fusion models with a joint loss function can further improve the prediction performance to 80.49% and 0.8331.

Evolutionary Learning-Derived Clinical-Radiomic Models for Predicting Early Recurrence of Hepatocellular Carcinoma after Resection

“…Furthermore, various deep learning-based radiomics methods have been proposed to design prediction models in medical applications [14]. The deep learning-based fusion model used clinical features and CECT images to predict early HCC recurrence [15]. However, the high-level radiomic features extracted by the convolutional layers may suffer from low medical interpretability and high overfitting probability, especially when the training dataset is not large enough, not conducive to understanding, and making diagnostic decisions.…”

CT images-based 3D convolutional neural network to predict early recurrence of solitary hepatocellular carcinoma after radical hepatectomy

“…They developed the clinical model (AUC = 0.7532) and the deep learning model (AUC = 0.7233) and proposed 4 fusion models to combine clinical data and deep learning features which achieved higher AUC. 21 However, they did not set up test group, which made it difficult to evaluate the generalization ability of the fusion models.…”

“… 17 , 18 The other is non-segmentation sampling (NSS) that does not segment the tumors region and retains all of the background information. 19 - 21 These two segmentation methods are widely used in studies but few studies have compared them.…”

Deep learning based radiomics for gastrointestinal cancer diagnosis and treatment: A minireview

“…Experimental results demonstrated that the integration of DL features and clinical features improved the prediction accuracy, and one combined model obtained the highest AUC of 0.825. The team improved their study by comparing the DL model with a conventional radiomics model, and one more combined model of another structure was added to the comparative analysis of their previous work[ 50 ]. The DL model performed better than the radiomics model with an average AUC of 0.7233 and accuracy of 69.52%, while one of the combined models surpassed the rest in the comparative analysis and reached 0.8248 and 78.66% in its average AUC and accuracy, respectively.…”