Variational Autoencoders for Generating Synthetic Tractography-Based Bundle Templates in a Low-Data Setting

Abstract

White matter tracts generated from whole brain tractography are often processed using automatic segmentation methods with standard atlases. Atlases are generated from hundreds of subjects, which becomes time-consuming to create and difficult to apply to all populations. In this study, we extended our prior work on using a deep generative model - Convolutional Variational Autoencoder - to map complex and data-intensive streamlines to a low-dimensional latent space given a limited sample size of 50 subjects from the ADNI3 dataset, to generate synthetic population-specific bundle templates using Kernel Density Estimation (KDE) on streamline embeddings. We conducted a quantitative shape analysis by calculating bundle shape metrics, and found that our bundle templates better capture the shape distribution of the bundles than the atlas data used in the original segmentation derived from young healthy adults. We further demonstrated the use of our framework for direct bundle segmentation from whole-brain tractograms.

Deep Normative Tractometry for Identifying Joint White Matter Macro- and Micro-structural Abnormalities in Alzheimer’s Disease

“…In addition, given the large number of streamlines and substantial proportion of false positives generated from tractography [14], modeling their joint distribution in a large cohort becomes challenging, but may potentially be tackled using deep learning methods. In our prior work [15], we showed that a convolutional variational autoencoder (ConvVAE) can embed tractography streamlines into a compact latent space and produce new bundles via generative sampling. Autoencoder-based architectures can also be used as a normative model (NM) [16], [17], where they encode statistical distributions of features from a healthy reference population.…”

Data augmentation: A comprehensive survey of modern approaches

“…Methods for synthetic data generation include explicitly creating samples with desired data distribution using computer graphics tools (e.g., in [2]) or algorithmically generating artificial data with the aid of special deep learning models (e.g., with techniques such as differential neural rendering [4], [5]. Neural style transfer [6], generative modeling techniques such as variational autoencoders (VAEs) [7] and generative adversarial networks (GANs) [8] have also been extensively used to generate synthetic data for training deep learning models. Another way to augment training data is by integrating existing data from several sources (e.g., in [9], [10], [11]).…”