AUTHOR=Zhang Yafang , Wei Qingyue , Huang Yini , Yao Zhao , Yan Cuiju , Zou Xuebin , Han Jing , Li Qing , Mao Rushuang , Liao Ying , Cao Lan , Lin Min , Zhou Xiaoshuang , Tang Xiaofeng , Hu Yixin , Li Lingling , Wang Yuanyuan , Yu Jinhua , Zhou Jianhua 

TITLE=Deep Learning of Liver Contrast-Enhanced Ultrasound to Predict Microvascular Invasion and Prognosis in Hepatocellular Carcinoma

JOURNAL=Frontiers in Oncology

VOLUME=12

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.878061

DOI=10.3389/fonc.2022.878061

ISSN=2234-943X

ABSTRACT=<sec><title>Background and Aims</title><p>Microvascular invasion (MVI) is a well-known risk factor for poor prognosis in hepatocellular carcinoma (HCC). This study aimed to develop a deep convolutional neural network (DCNN) model based on contrast-enhanced ultrasound (CEUS) to predict MVI, and thus to predict prognosis in patients with HCC.</p></sec><sec><title>Methods</title><p>A total of 436 patients with surgically resected HCC who underwent preoperative CEUS were retrospectively enrolled. Patients were divided into training (<italic>n</italic> = 301), validation (<italic>n</italic> = 102), and test (<italic>n</italic> = 33) sets. A clinical model (Clinical model), a CEUS video-based DCNN model (CEUS-DCNN model), and a fusion model based on CEUS video and clinical variables (CECL-DCNN model) were built to predict MVI. Survival analysis was used to evaluate the clinical performance of the predicted MVI.</p></sec><sec><title>Results</title><p>Compared with the Clinical model, the CEUS-DCNN model exhibited similar sensitivity, but higher specificity (71.4% vs. 38.1%, <italic>p</italic> = 0.03) in the test group. The CECL-DCNN model showed significantly higher specificity (81.0% vs. 38.1%, <italic>p</italic> = 0.005) and accuracy (78.8% vs. 51.5%, <italic>p</italic> = 0.009) than the Clinical model, with an AUC of 0.865. The Clinical predicted MVI could not significantly distinguish OS or RFS (both <italic>p</italic> &gt; 0.05), while the CEUS-DCNN predicted MVI could only predict the earlier recurrence (hazard ratio [HR] with 95% confidence interval [CI 2.92 [1.1–7.75], <italic>p</italic> = 0.024). However, the CECL-DCNN predicted MVI was a significant prognostic factor for both OS (HR with 95% CI: 6.03 [1.7–21.39], <italic>p</italic> = 0.009) and RFS (HR with 95% CI: 3.3 [1.23–8.91], <italic>p</italic> = 0.011) in the test group.</p></sec><sec><title>Conclusions</title><p>The proposed CECL-DCNN model based on preoperative CEUS video can serve as a noninvasive tool to predict MVI status in HCC, thereby predicting poor prognosis.</p></sec>