Huang Zhenhuan, Huang Wanrong, Jiang Lu, Zheng Yao, Pan Yifan, Yan Chuan, Ye Rongping, Weng Shuping, Li Yueming
Department of Radiology, The First Affiliated Hospital of Fujian Medical University, Fuzhou, Fujian 350005, China (Z.H., W.H., L.J., Y.Z., Y.P., C.Y., R.Y., Y.L.); Department of Radiology, Longyan First Affiliated Hospital of Fujian Medical University, Longyan, Fujian 364000, China (Z.H.).
Department of Radiology, The First Affiliated Hospital of Fujian Medical University, Fuzhou, Fujian 350005, China (Z.H., W.H., L.J., Y.Z., Y.P., C.Y., R.Y., Y.L.).
Acad Radiol. 2025 Apr;32(4):1971-1980. doi: 10.1016/j.acra.2024.10.007. Epub 2024 Oct 28.
Accurate prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) is crucial for guiding treatment. This study evaluates and compares the performance of clinicoradiologic, traditional radiomics, deep-learning radiomics, feature fusion, and decision fusion models based on multi-region MR habitat imaging using six machine-learning classifiers.
We retrospectively included 300 HCC patients. The intratumoral and peritumoral regions were segmented into distinct habitats, from which radiomics and deep-learning features were extracted using arterial phase MR images. To reduce feature dimensionality, we applied intra-class correlation coefficient (ICC) analysis, Pearson correlation coefficient (PCC) filtering, and recursive feature elimination (RFE). Based on the selected optimal features, prediction models were constructed using decision tree (DT), K-nearest neighbors (KNN), logistic regression (LR), random forest (RF), support vector machine (SVM), and XGBoost (XGB) classifiers. Additionally, fusion models were developed utilizing both feature fusion and decision fusion strategies. The performance of these models was validated using the area under the receiver operating characteristic curve (ROC AUC), calibration curves, and decision curve analysis.
The decision fusion model (VOI-Peri10-1) using LR and integrating clinicoradiologic, radiomics, and deep-learning features achieved the highest AUC of 0.808 (95% confidence interval [CI]: 0.807-0.912) in the test cohort, with good calibration (Hosmer-Lemeshow test, P > 0.050) and clinical net benefit.
The LR-based decision fusion model integrating clinicoradiologic, radiomics, and deep-learning features shows promise for preoperative prediction of MVI in HCC, aiding in patient outcome predictions and personalized treatment planning.
准确预测肝细胞癌（HCC）中的微血管侵犯（MVI）对于指导治疗至关重要。本研究使用六种机器学习分类器，评估并比较基于多区域MR栖息地成像的临床放射学、传统放射组学、深度学习放射组学、特征融合和决策融合模型的性能。
我们回顾性纳入了300例HCC患者。将肿瘤内和肿瘤周围区域分割为不同的栖息地，使用动脉期MR图像从中提取放射组学和深度学习特征。为了降低特征维度，我们应用了组内相关系数（ICC）分析、Pearson相关系数（PCC）滤波和递归特征消除（RFE）。基于选定的最优特征，使用决策树（DT）、K近邻（KNN）、逻辑回归（LR）、随机森林（RF）、支持向量机（SVM）和XGBoost（XGB）分类器构建预测模型。此外，利用特征融合和决策融合策略开发了融合模型。使用受试者工作特征曲线下面积（ROC AUC）、校准曲线和决策曲线分析来验证这些模型的性能。
在测试队列中，使用LR并整合临床放射学、放射组学和深度学习特征的决策融合模型（VOI-Peri10-1）实现了最高的AUC，为0.808（95%置信区间[CI]：0.807-0.912），具有良好的校准（Hosmer-Lemeshow检验，P>0.050）和临床净效益。
整合临床放射学、放射组学和深度学习特征的基于LR的决策融合模型在术前预测HCC中的MVI方面显示出前景，有助于患者预后预测和个性化治疗规划。
