Department of Radiology, The First Affiliated Hospital of Jinan University, Guangzhou, Guangdong, China.
Department of MR, Zhongshan City People's Hospital Affiliated to Sun Yat-sen University, Zhongshan, Guangdong, China.
J Magn Reson Imaging. 2021 Jul;54(1):134-143. doi: 10.1002/jmri.27538. Epub 2021 Feb 8.
Microvascular invasion (MVI) is a critical prognostic factor of hepatocellular carcinoma (HCC). However, it could only be obtained by postoperative histological examination.
To develop an end-to-end deep-learning models based on MRI images for preoperative prediction of MVI in HCC patients who underwent surgical resection.
Retrospective.
Two hundred and thirty-seven patients with histologically confirmed HCC.
1.5 T and 3.0 T.
Axial T -weighted (T -w) with turbo spin echo sequence, T -Spectral Presaturation with Inversion Recovery (T -SPIR), and dynamic contrast-enhanced (DCE) imaging with fat suppressed enhanced T high-resolution isotropic volume examination.
The patients were randomly divided into training (N = 158) and validation (N = 79) sets. Data augmentation by random rotation was performed on the training set and the sample size increased to 1940 for each MR sequence. A three-dimensional convolutional neural network (3D CNN) was used to develop four deep-learning models, including three single-layer models based on single-sequence, and fusion model combining three sequences. MVI status was obtained from the postoperative pathology reports.
The dice similarity coefficient (DSC) and Hausdorff distance (HD) were applied to assess the similarity and reproducibility between the manual segmentations of tumor from two radiologists. Receiver operating characteristic curve analysis was used to evaluate model performance. MVI was identified in 92 (38.8%) patients. Good reproducibility with interobserver DSCs of 0.90, 0.89, and 0.89 and HDs of 4.09, 3.67, and 3.60 was observed for PVP, T WI, and T -SPIR, respectively. The fusion model achieved an area under the curve (AUC) of 0.81, sensitivity of 69%, and specificity of 79% in the training set and 0.72, sensitivity of 55%, and specificity of 81% in the validation set.
3D CNN model may serve as a noninvasive tool to predict MVI in HCC, whereas its accuracy needs to be enhanced with larger cohort.
3 TECHNICAL EFFICACY: Stage 2.
微血管侵犯（MVI）是肝细胞癌（HCC）的一个关键预后因素。然而，它只能通过术后组织学检查获得。
开发一种基于 MRI 图像的端到端深度学习模型，用于预测接受手术切除的 HCC 患者的 MVI。
回顾性。
237 名经组织学证实的 HCC 患者。
1.5T 和 3.0T。
轴位 T 加权（T - w）涡轮自旋回波序列、T - 谱预饱和反转恢复（T - SPIR）和动态对比增强（DCE）成像，脂肪抑制增强 T 高分辨率各向同性容积检查。
患者随机分为训练集（N=158）和验证集（N=79）。在训练集上进行随机旋转数据增强，每个 MR 序列的样本量增加到 1940。使用三维卷积神经网络（3D CNN）开发了四个深度学习模型，包括三个基于单序列的单层模型和融合三个序列的融合模型。MVI 状态从术后病理报告中获得。
采用 Dice 相似系数（DSC）和 Hausdorff 距离（HD）评估两位放射科医生手动分割肿瘤的相似性和可重复性。使用受试者工作特征曲线分析评估模型性能。在 92 例（38.8%）患者中发现 MVI。PVP、T WI 和 T - SPIR 的观察者间 DSC 分别为 0.90、0.89 和 0.89，HD 分别为 4.09、3.67 和 3.60，具有良好的可重复性。融合模型在训练集的 AUC 为 0.81，敏感性为 69%，特异性为 79%，在验证集的 AUC 为 0.72，敏感性为 55%，特异性为 81%。
3D CNN 模型可能是一种预测 HCC 患者 MVI 的非侵入性工具，但需要更大的队列来提高其准确性。
3 技术功效：2 级。
