Preoperative identification of microvascular invasion in hepatocellular carcinoma by XGBoost and deep learning

  • PDF / 2,679,333 Bytes
  • 13 Pages / 595.276 x 790.866 pts Page_size
  • 42 Downloads / 182 Views

DOWNLOAD

REPORT


ORIGINAL ARTICLE – CLINICAL ONCOLOGY

Preoperative identification of microvascular invasion in hepatocellular carcinoma by XGBoost and deep learning Yi‑Quan Jiang1 · Su‑E Cao2 · Shilei Cao3 · Jian‑Ning Chen4 · Guo‑Ying Wang1 · Wen‑Qi Shi2 · Yi‑Nan Deng1 · Na Cheng4 · Kai Ma3 · Kai‑Ning Zeng1 · Xi‑Jing Yan1 · Hao‑Zhen Yang5 · Wen‑Jing Huan5 · Wei‑Min Tang5 · Yefeng Zheng3 · Chun‑Kui Shao4 · Jin Wang2 · Yang Yang1 · Gui‑Hua Chen6,7  Received: 30 June 2020 / Accepted: 18 August 2020 © The Author(s) 2020

Abstract Purpose  Microvascular invasion (MVI) is a valuable predictor of survival in hepatocellular carcinoma (HCC) patients. This study developed predictive models using eXtreme Gradient Boosting (XGBoost) and deep learning based on CT images to predict MVI preoperatively. Methods  In total, 405 patients were included. A total of 7302 radiomic features and 17 radiological features were extracted by a radiomics feature extraction package and radiologists, respectively. We developed a XGBoost model based on radiomics features, radiological features and clinical variables and a three-dimensional convolutional neural network (3D-CNN) to predict MVI status. Next, we compared the efficacy of the two models. Results  Of the 405 patients, 220 (54.3%) were MVI positive, and 185 (45.7%) were MVI negative. The areas under the receiver operating characteristic curves (AUROCs) of the Radiomics-Radiological-Clinical (RRC) Model and 3D-CNN Model in the training set were 0.952 (95% confidence interval (CI) 0.923–0.973) and 0.980 (95% CI 0.959–0.993), respectively (p = 0.14). The AUROCs of the RRC Model and 3D-CNN Model in the validation set were 0.887 (95% CI 0.797–0.947) and 0.906 (95% CI 0.821–0.960), respectively (p = 0.83). Based on the MVI status predicted by the RRC and 3D-CNN Models, the mean recurrence-free survival (RFS) was significantly better in the predicted MVI-negative group than that in the predicted MVI-positive group (RRC Model: 69.95 vs. 24.80 months, p  5 cm vs. ≤ 5 cm); (6) tumor growth pattern (intrahepatic growth vs. extrahepatic growth); (7) intratumoral hemorrhage (absence vs. presence); (8) intratumoral necrosis (absence vs. presence); (9) tumor margin (smooth vs. nonsmooth); (10) enhancing “capsule” (absence vs. presence); (11) AP hyperenhancement (absence vs. presence); (12) internal arteries (absence vs. presence); (13) peritumoral enhancement (absence vs. presence); (14) mosaic pattern or nodulein-nodule pattern (absence vs. presence); (15) nonperipheral washout (absence vs. presence); (16) hypodense halos (absence vs. presence); and (17) tumor steatosis (absence vs. presence). The largest nodule was evaluated if multiple nodules existed. The definitions of some radiological features are shown in Supplemental methods. Clinical variables Baseline data of the patients including age, sex, background liver disease, diabetes, surgery type, primary tumor size, tumor count, α-fetoprotein (AFP) level, aspartate aminotransferase (AST) level, alanine aminotransferase (ALT) level, prothrombin time (PT), internati