Filtering in tractography using autoencoders (FINTA)

Current brain white matter fiber tracking techniques show a number of problems, including: generating large proportions streamlines that do not accurately describe the underlying anatomy; extracting are supported by diffusion signal; and under-representing some populations, among others. In this paper, we novel autoencoder-based learning method to filter from MRI tractography, hence, obtain more reliable tractograms. Our method, dubbed FINTA (Filtering in Tractography using Autoencoders) uses raw, unlabeled tractograms train autoencoder, learn robust representation streamlines. Such an embedding is then used undesired streamline samples nearest neighbor algorithm. experiments on both synthetic vivo human tractography data accuracy scores exceeding 90\% threshold test set. Results reveal has superior filtering performance compared conventional, anatomy-based methods, RecoBundles state-of-the-art method. Additionally, demonstrate can be applied partial without requiring changes framework. We also proposed generalizes well across different methods datasets, shortens significantly computation time for (>1 M streamlines) Together, work brings forward new deep framework based autoencoders, which offers flexible powerful bundling could enhance tractometry connectivity analyses.