FIESTA: Autoencoders for accurate fiber segmentation in tractography

Félix Dumais, Jon Haitz Legarreta, Carl Lemaire, Philippe Poulin, François Rheault, Laurent Petit, Muhamed Barakovic, Stefano Magon, Maxime Descoteaux, Pierre-Marc Jodoin

Abstract: Highlights•FIESTA is an autoencoder-based pipeline for dMRI tractography segmentation.•FIESTA is more reliable than current state-of-the-art automatic bundling methods.•FIESTA can recover hard-to-track streamlines thanks to its generative sampling module.•Contrastive learning based on QuickBundlesX clusters can be used to build a useful autoencoder’s latent representation of whole-brain tractograms.•FIESTA bundle definitions are easily editable with no need to re-train a neural network.