[Seminar] Manifold-Aware CycleGAN for High-Resolution Structural-to-DTI Synthesis

Summary: Unpaired image-to-image translation has been applied successfully to natural images but has received very little attention for manifold-valued data such as in diffusion tensor imaging (DTI). The non-Euclidean nature of DTI prevents current generative adversarial networks (GANs) from generating plausible images and has mainly limited their application to diffusion MRI scalar maps, such as fractional anisotropy (FA) or mean diffusivity (MD). Even if these scalar maps are clinically useful, they mostly ignore fiber orientations and therefore have limited applications for analyzing brain fibers. In this presentation, a Manifold-Aware Wassertein CycleGAN (MAWCGAN) leveraging the Log-Euclidean metric and high-resolution structural images (T1w) to generate high-resolution DTIs will be presented. We demonstrate that high-resolution T1w images can be used to increase the spatial resolution of DTIs and that considering the underlying data manifold geometry helps generating plausible data on such manifold.