Abstracts

Rationale: Convolutional neural networks (CNNs) automatically detect patterns of interest in images and have demonstrated image-classification performance at or above the level of humans. This study aims to develop and test a deep learning algorithm to automatically detect cortical tubers in magnetic resonance imaging (MRI) and to explore the utility of deep learning in rare disorders with limited data. Methods: T2 and FLAIR axial images with and without tubers were extracted from MRIs of patients with tuberous sclerosis complex (TSC) and controls, respectively. We performed data augmentation of images in the training set to better train the CNNs and make them more resistant to noise. Data augmentation of images was not performed in the validation or test set. We trained three different CNN architectures on a training dataset (TSCCNN, InceptionV3, and ResNet50) and selected the one with the lowest binary cross-entropy loss in the validation dataset. The best CNN was evaluated on the test dataset. We visualized the most relevant regions for classification with gradient-weighted class activation maps (Grad-CAM) and saliency maps. Results: The TSC patients and controls included in the study were similar, except TSC patients were younger than controls at the time of imaging (median 9.5 years versus 12.4 years). The data were divided into 566 images for TSC training (69 patients), 130 images for TSC validation (20 patients), 210 images for TSC testing (25 patients), 561 images for control training (69 patients), 118 images for control validation (20 patients), and 226 images for control testing (25 patients). The training file, merging the TSC and control training images and after data augmentation contained 5,634 images. The InceptionV3 CNN architecture performed best in the validation set and was evaluated in the test set with the following results: sensitivity: 0.95, specificity: 0.95, positive predictive value: 0.94, negative predictive value: 0.95, F1-score: 0.95, accuracy: 0.95, and area under the curve: 0.99 (Table 1). Grad-CAM and saliency maps showed that tubers reside in the most relevant regions for classification within each image (Figure 1). Conclusions: This study shows that deep learning algorithms are able to detect tubers in MRI, and deep learning can be applied to a small dataset in a rare neurological disorder. Funding: No funding