Improving automated multiple sclerosis lesion segmentation with a cascaded 3D convolutional neural network approach