School of Automation, Harbin University of Science and Technology, Heilongjiang, Harbin, 150080, China; CAS Key Laboratory of Molecular Imaging, Institute of Automation, Chinese Academy of Sciences, Beijing, 100190, China; Beijing Advanced Innovation Center for Big Data-Based Precision Medicine, School of Medicine, Beihang University, Beijing, 100191, China.
School of Automation, Harbin University of Science and Technology, Heilongjiang, Harbin, 150080, China.
Eur J Radiol. 2019 Jul;116:128-134. doi: 10.1016/j.ejrad.2019.04.022. Epub 2019 May 1.
To noninvasively differentiate meningioma grades by deep learning radiomics (DLR) model based on routine post-contrast MRI.
We enrolled 181 patients with histopathologic diagnosis of meningioma who received post-contrast MRI preoperative examinations from 2 hospitals (99 in the primary cohort and 82 in the validation cohort). All the tumors were segmented based on post-contrast axial T1 weighted images (T1WI), from which 2048 deep learning features were extracted by the convolutional neural network. The random forest algorithm was used to select features with importance values over 0.001, upon which a deep learning signature was built by a linear discriminant analysis classifier. The performance of our DLR model was assessed by discrimination and calibration in the independent validation cohort. For comparison, a radiomic model based on hand-crafted features and a fusion model were built.
The DLR signature comprised 39 deep learning features and showed good discrimination performance in both the primary and validation cohorts. The area under curve (AUC), sensitivity, and specificity for predicting meningioma grades were 0.811(95% CI, 0.635-0.986), 0.769, and 0.898 respectively in the validation cohort. DLR performance was superior over the hand-crafted features. Calibration curves of DLR model showed good agreements between the prediction probability and the observed outcome of high-grade meningioma.
Using routine MRI data, we developed a DLR model with good performance for noninvasively individualized prediction of meningioma grades, which achieved a quantization capability superior over the hand-crafted features. This model has potential to guide and facilitate the clinical decision-making of whether to observe or to treat patients by providing prognostic information.
通过基于常规对比后 MRI 的深度学习放射组学（DLR）模型，对脑膜瘤分级进行无创区分。
我们纳入了 2 家医院（原发队列 99 例，验证队列 82 例）经病理组织学诊断为脑膜瘤并接受术前对比后 MRI 检查的 181 例患者。所有肿瘤均基于对比后轴位 T1 加权图像（T1WI）进行分割，由卷积神经网络提取 2048 个深度学习特征。随机森林算法用于选择重要值大于 0.001 的特征，在此基础上通过线性判别分析分类器构建深度学习特征签名。通过独立验证队列评估我们的 DLR 模型的判别和校准性能。为了比较，还构建了基于手工特征的放射组学模型和融合模型。
DLR 特征签名包含 39 个深度学习特征，在原发队列和验证队列中均表现出良好的判别性能。验证队列中，预测脑膜瘤分级的曲线下面积（AUC）、敏感性和特异性分别为 0.811（95%CI，0.635-0.986）、0.769 和 0.898。DLR 性能优于手工特征。DLR 模型的校准曲线显示，预测高级别脑膜瘤的概率与观察结果之间具有良好的一致性。
使用常规 MRI 数据，我们开发了一种 DLR 模型，该模型具有良好的脑膜瘤分级无创个体化预测性能，其量化能力优于手工特征。该模型有可能通过提供预后信息，为是否观察或治疗患者提供决策指导，从而辅助临床决策。
