Fig. 2From: Convolutional neural network-based magnetic resonance image differentiation of filum terminale ependymomas from schwannomasPipeline of the proposed diagnostic system. a training images. b data augmentation consisted of random rotation, random horizontal-flip and normalization. c five-fold split of cross validation. d for every mode, five CNN models are trained at image level. e integrated diagnostic model from five CNN models. f external test images of one case. g probability of model diagnosis from 3 MRI modalities. h output of diagnostic systemBack to article page