Multiscale Metamorphic VAE for 3D Brain MRI Synthesis

2023/01/09 by Jaivardhan Kapoor, Jakob H. Macke, Kapoor, Jaivardhan +3 · 4 citations
Computer Science · Medicine · Biochemistry, Genetics and Molecular Biology · #Generative Adversarial Networks and Image Synthesis #Advanced Neuroimaging Techniques and Applications #Cell Image Analysis Techniques

paper · pdf · doi:10.48550/arxiv.2301.03588

Abstract

Generative modeling of 3D brain MRIs presents difficulties in achieving high visual fidelity while ensuring sufficient coverage of the data distribution. In this work, we propose to address this challenge with composable, multiscale morphological transformations in a variational autoencoder (VAE) framework. These transformations are applied to a chosen reference brain image to generate MRI volumes, equipping the model with strong anatomical inductive biases. We structure the VAE latent space in a way such that the model covers the data distribution sufficiently well. We show substantial performance improvements in FID while retaining comparable, or superior, reconstruction quality compared to prior work based on VAEs and generative adversarial networks (GANs).

Cited by

• From Healthy Scans to Annotated Tumors: A Tumor Fabrication Framework for 3D Brain MRI Synthesis
• MRI Super-Resolution with Deep Learning: A Comprehensive Survey
• MRExtrap: Longitudinal Aging of Brain MRIs using Linear Modeling in Latent Space
• MInDI-3D: Iterative Deep Learning in 3D for Sparse-view Cone Beam Computed Tomography

Related

• Generating Novel Brain Morphology by Deforming Learned Templates
• Generation of 3D Brain MRI Using Auto-Encoding Generative Adversarial Networks
• Neuromorphologicaly-preserving Volumetric data encoding using VQ-VAE
• Make-A-Volume: Leveraging Latent Diffusion Models for Cross-Modality 3D Brain MRI Synthesis
• Modelling the Distribution of 3D Brain MRI using a 2D Slice VAE