Development and Validation of a Deep Learning–Based Model to Distinguish Glioblastoma from Solitary Brain Metastasis Using Conventional MR Images

Abstract

BACKGROUND AND PURPOSE: Differentiating glioblastoma from solitary brain metastasis preoperatively using conventional MR images is challenging. Deep learning models have shown promise in performing classification tasks. The diagnostic performance of a deep learning-based model in discriminating glioblastoma from solitary brain metastasis using preoperative conventional MR images was evaluated. MATERIALS AND METHODS: Records of 598 patients with histologically confirmed glioblastoma or solitary brain metastasis at our institution between February 2006 and December 2017 were retrospectively reviewed. Preoperative contrastenhanced T1WI and T2WI were preprocessed and roughly segmented with rectangular regions of interest. A deep neural network was trained and validated using MR images from 498 patients. The MR images of the remaining 100 were used as an internal test set. An additional 143 patients from another tertiary hospital were used as an external test set. The classifications of ResNet-50 and 2 neuroradiologists were compared for their accuracy, precision, recall, F1 score, and area under the curve. RESULTS: The areas under the curve of ResNet-50 were 0.889 and 0.835 in the internal and external test sets, respectively. The area under the curve of neuroradiologists 1 and 2 were 0.889 and 0.768 in the internal test set and 0.857 and 0.708 in the external test set, respectively. CONCLUSIONS: A deep learning-based model may be a supportive tool for preoperative discrimination between glioblastoma and solitary brain metastasis using conventional MR images. ABBREVIATIONS: CE ¼ contrast enhanced; GBM ¼ glioblastoma; ROC ¼ receiver operating characteristic; AUC ¼ area under the curve

Machine Learning Applications for Differentiation of Glioma from Brain Metastasis—A Systematic Review

“…Cross-validation, particularly leave-one-out cross validation (LOOCV) (n = 9), was performed in 21 studies. Two studies [31,36] presented a three-way split of their dataset into training, validation, and testing set. External validation sets stem from a geographically distinct location and should ensure that the model generalizes well onto data from foreign populations.…”

Neuroinflammation and immunoregulation in glioblastoma and brain metastases: Recent developments in imaging approaches

“…data to improve the differentiation between GBM and brain metastases [74][75][76][77][78]. Novel diagnostic support systems based on radiomic features extracted from post-contrast 3 diffusion tensor imaging (DT1) MR images may help improving the distinction between solitary brain metastases and GBM with high diagnosis performance and generalizability [74].…”

“…Novel diagnostic support systems based on radiomic features extracted from post-contrast 3 diffusion tensor imaging (DT1) MR images may help improving the distinction between solitary brain metastases and GBM with high diagnosis performance and generalizability [74]. Machine learning and deep learning-based models applied to conventional MR images may support pre-operative discrimination between GBM and solitary brain metastasis [75][76][77], and deep learning network models that allow automated, on-site analysis of resected tumor specimens based on confocal laser endoscopic techniques image data sets have been developed [78]. Other parameters, such as the cerebral blood volume gradient in the peritumoral brain zone, may enable the differentiation of GBMs from metastases [79].…”

Brain tumours, brain metastases, and neuroinflammation: Insights from neuroimaging studies.

“…Computational-aided quantitative analysis of MRI images may improve the accuracy in differentiating GBM from metastases, and texture features are more significant than fractal-based features for that purpose [76]. Increasingly, machine learning algorithms have been applied to imaging data to improve the differentiation between GBM and brain metastases [77][78][79][80][81]. Novel diagnostic support systems based on radiomic features extracted from post-contrast 3DT1 MR images may help improving the distinction between solitary brain metastases and GBM with high diagnosis performance and generalizability [77].…”

“…Novel diagnostic support systems based on radiomic features extracted from post-contrast 3DT1 MR images may help improving the distinction between solitary brain metastases and GBM with high diagnosis performance and generalizability [77]. Machine learning and deep learning-based models applied to conventional MR images may support preoperative discrimination between GBM and solitary brain metastasis conventional MR images [78][79][80], and deep learning network models that allow automated, on-site analysis of resected tumor specimens based on confocal laser endoscopic techniques image datasets have been developed [81]. Other parameters such as the cerebral blood volume gradient in the peritumoral brain zone may enable the differentiation of GBMs from metastases [82].…”