Self-Supervised Slice-to-Volume Reconstruction with Gaussian Representations for Fetal MRI

TL;DR

GaussianSVR tackles the challenge of reconstructing high-fidelity 3D fetal brain volumes from motion-corrupted 2D slices without ground-truth volumes by representing the volume with 3D Gaussian primitives and training via a simulated forward slice acquisition model. The method employs a self-supervised, multi-resolution optimization that jointly estimates Gaussian parameters and slice-wise transformations, using a loss that combines L1 data fidelity, D-SSIM, and TV regularization. Key contributions include the first application of 3D Gaussian primitives for SVR and a scalable, self-supervised training regime that outperforms baselines on FeTA data, with quantitative gains in PSNR and SSIM. The approach reduces dependence on ground-truth data and offers a computationally efficient, resolution-agnostic pathway for fetal MRI reconstruction.

Abstract

Reconstructing 3D fetal MR volumes from motion-corrupted stacks of 2D slices is a crucial and challenging task. Conventional slice-to-volume reconstruction (SVR) methods are time-consuming and require multiple orthogonal stacks for reconstruction. While learning-based SVR approaches have significantly reduced the time required at the inference stage, they heavily rely on ground truth information for training, which is inaccessible in practice. To address these challenges, we propose GaussianSVR, a self-supervised framework for slice-to-volume reconstruction. GaussianSVR represents the target volume using 3D Gaussian representations to achieve high-fidelity reconstruction. It leverages a simulated forward slice acquisition model to enable self-supervised training, alleviating the need for ground-truth volumes. Furthermore, to enhance both accuracy and efficiency, we introduce a multi-resolution training strategy that jointly optimizes Gaussian parameters and spatial transformations across different resolution levels. Experiments show that GaussianSVR outperforms the baseline methods on fetal MR volumetric reconstruction. Code will be available upon acceptance.

Figures (2)

• Figure 1: Overview of the framework of the proposed GaussianSVR. Solid lines indicate forward propagation; dashed lines indicate backward propagation.
• Figure 2: Qualitative volumetric reconstruction results on a single subject from motion-corrupted scans on the FeTA dataset.