AUTHOR=Zhang Jian , Huang Shenglan , Xu Yongkang , Wu Jianbing 

TITLE=Diagnostic Accuracy of Artificial Intelligence Based on Imaging Data for Preoperative Prediction of Microvascular Invasion in Hepatocellular Carcinoma: A Systematic Review and Meta-Analysis

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.763842

DOI=10.3389/fonc.2022.763842

ISSN=2234-943X

ABSTRACT=Background: The presence of microvascular invasion (MVI) is considered as an independent prognostic factor associated with the early recurrence and poor survival after resection in hepatocellular carcinoma (HCC) patients. Artificial intelligence, mainly consist of machine learning algorithms (MLAs) and deep learning algorithms (DLAs) have been widely used to MVI prediction in medical imaging.
Aim: To assess the diagnostic accuracy of artificial intelligence algorithms for non-invasive predication of MVI preoperatively based on imaging data. 
Methods: Original studies reporting artificial intelligence algorithms for non-invasive predication of MVI preoperatively based on quantitative imaging data were identified in the databases PubMed, Embase and Web of Science. The qualities of the included studies were assessed according to QUADAS-2 scale. The pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR) with their 95% confidence intervals (95% CIs) were calculated using a random effects model. A summary receiver operating characteristic curve and the area under the curve was generated to assess the diagnostic accuracy of the deep learning models or machine learning models.In machine learning group, we further performed meta-regression and subgroup analyses to identify the source of heterogeneity.
Results: Data of 16 included studies with 4759 cases are available for meta-analysis. 4 studies reporting deep learning models,12 studies reporting machine learning models and 2 studies compared the efficiency of deep learning models and machine learning models. For predicative performance of deep learning models, the pooled sensitivity, specificity, PLR, NLR and AUC value were 0.84[0.75-0.90], 0.84[0.77-0.89], 5.14[3.53-7.48],0.2[0.12-0.31] and 0.90[0.87-0.93], respectively; for machine learning models, 0.77[0.71-0.82],0.77[0.73-0.80],3.30[2.83-3.84], 0.30[0.24-0.38] and 0.82[0.79-0.85], respectively. Subgroup analyses showed that significant difference between solitary tumor subgroup and multiple tumor subgroup was observed in the pooled sensitivity, NLR and AUC.
Conclusion: This meta-analysis demonstrates the promising potential of machine learning and deep learning method for MVI status prediction and further applying in clinical decisions making. Deep learning models rather than machine learning models have more excellent performance, in term of accuracy of MVI prediction, methodology and cost effectiveness.