AUTHOR=Liu Weiwei , Zhang Lifan , Xin Zhaodan , Zhang Haili , You Liting , Bai Ling , Zhou Juan , Ying Binwu 

TITLE=A Promising Preoperative Prediction Model for Microvascular Invasion in Hepatocellular Carcinoma Based on an Extreme Gradient Boosting Algorithm

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.852736

DOI=10.3389/fonc.2022.852736

ISSN=2234-943X

ABSTRACT=Background: The non-invasive preoperative diagnosis of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) is extremely vital to precise surgical decision making and prognosis of patients. In this study, we aimed to develop a MVI prediction model with valid performance and clinical interpretability.
Methods: 2160 HCC patients without macroscopic invasion who underwent hepatectomy for the first time in West China Hospital from January 2015 to June 2019 were retrospectively included, and these patients were randomly divided into a training cohort and a validation cohort in the ratio of 8:2. Preoperative demographic features, imaging characteristics and laboratory indexes of patients were collected. Five machine learning algorithms were used including Logistic Regression, Random Forest (RF), Support Vector Machine (SVM), extreme gradient boosting (XGBoost), and Multi-Layer Perception (MLP), and performance was evaluated using the area under the receiver operating characteristic curve (AUC). We also draw the SHAP value to explain the influence of each feature on the MVI prediction model.
Results: The top six important preoperative factors associated with MVI were maximum image diameter, protein induced by vitamin K absence or antagonist-II (PIVKA-II), α-fetoprotein (AFP) level, satellite ovens, alanine aminotransferase (AST) / aspartate aminotransferase (ALT) and alanine aminotransferase (AST) according to the XGBoost model. The XGBoost model for preoperative prediction of MVI had exhibited better AUC (0.8, 95% CI: 0.74-0.83) than other prediction model. Furthermore, to facilitate the utilizing of the model in clinical settings, we developed a user-friendly online calculator for MVI risk prediction based on the XGBoost model.
Conclusion: The XGBoost model achieved outstanding performance for non-invasive preoperative prediction of MVI based on a big data. Moreover, the MVI risk calculator would assist clinicians to conveniently determine the optimal therapeutic remedy, ameliorating the prognosis of HCC patients.