Parcellation of whole brain tractogram is a critical step to study brain white matter structures and connectivity patterns. The existing methods based on supervised classification of streamlines into predefined streamline bundle types are not designed to explore sub-bundle structures, and methods with manually designed features are expensive to compute streamline-wise similarities. To resolve these issues, we proposed a novel atlas-free method that learnt a latent space using a deep recurrent autoencoder which efficiently embedded any lengths of streamlines to fixed-size feature vectors, namely, streamline embeddings, and enabled tractogram parcellation via unsupervised clustering in the latent space. The method is evaluated on the ISMRM 2015 tractography challenge dataset, and shows the ability to discriminate major bundles with unsupervised clustering and query streamline based on similarity. The learnt latent representations of streamlines and bundles also open the possibility of quantitatively studying any granularities of sub-bundle structures with generic data mining techniques.
Switch-Related and General Preparation Processes in Task-Switching: Evidence from Multivariate Pattern Classification of EEG Data
