An unsupervised 3D segmentation autoencoder (SegAE) using fully convolutional layers on three scales. The network is optimized based on Cosine proximity; and outputs four channels for GM, WM, CSF, and WMH separately.
Full description of this method.

Grand Challenge
An unsupervised 3D segmentation autoencoder (SegAE) using fully convolutional layers on three scales. The network is optimized based on Cosine proximity; and outputs four channels for GM, WM, CSF, and WMH separately.
Full description of this method.