Department of Hepatic Surgery and Liver Transplantation Center, The Third Affiliated Hospital, Organ Transplantation Institute, Sun Yat-Sen University, 600 Tianhe Road, Guangzhou, 510630, Guangdong, China.
Department of Radiology, The Third Affiliated Hospital, Sun Yat-Sen University, 600 Tianhe Road, Guangzhou, 510630, China.
J Cancer Res Clin Oncol. 2021 Mar;147(3):821-833. doi: 10.1007/s00432-020-03366-9. Epub 2020 Aug 27.
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.
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.
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 < 0.001; 3D-CNN Model: 64.06 vs. 31.05 months, p = 0.027).
The RRC Model and 3D-CNN models showed considerable efficacy in identifying MVI preoperatively. These machine learning models may facilitate decision-making in HCC treatment but requires further validation.
微血管侵犯（MVI）是肝细胞癌（HCC）患者生存的有价值的预测因子。本研究基于 CT 图像，使用极端梯度提升（XGBoost）和深度学习开发了预测 MVI 的术前预测模型。
共纳入 405 例患者。通过放射组学特征提取包和放射科医生分别提取了 7302 个放射组学特征和 17 个放射学特征。我们开发了一个基于放射组学特征、放射学特征和临床变量的 XGBoost 模型和三维卷积神经网络（3D-CNN）来预测 MVI 状态。然后，我们比较了两种模型的疗效。
在 405 例患者中，220 例（54.3%）为 MVI 阳性，185 例（45.7%）为 MVI 阴性。训练集中，放射组学-放射学-临床（RRC）模型和 3D-CNN 模型的受试者工作特征曲线（AUROC）下面积分别为 0.952（95%置信区间（CI）0.923-0.973）和 0.980（95%CI 0.959-0.993）（p=0.14）。验证集中，RRC 模型和 3D-CNN 模型的 AUROC 分别为 0.887（95%CI 0.797-0.947）和 0.906（95%CI 0.821-0.960）（p=0.83）。基于 RRC 和 3D-CNN 模型预测的 MVI 状态，预测 MVI 阴性组的平均无复发生存（RFS）明显优于预测 MVI 阳性组（RRC 模型：69.95 个月比 24.80 个月，p<0.001；3D-CNN 模型：64.06 个月比 31.05 个月，p=0.027）。
RRC 模型和 3D-CNN 模型在术前识别 MVI 方面具有相当的疗效。这些机器学习模型可能有助于 HCC 治疗决策，但需要进一步验证。
