Potential and challenges of generative adversarial networks for super-resolution in 4D Flow MRI

Time-resolved three-dimensional phase-contrast MRI (4D Flow MRI) enables non-invasive quantification of blood flow and derivation of hemodynamic parameters. However, its clinical application is limited by low spatial resolution and noise, particularly a ffecting velocity measurements near vessel walls. Machine learning-based super-resolution has shown promise in addressing these limitations, but challenges remain, not least in recovering near-wall velocities. Generative adversarial networks (GANs) o ff er a compelling solution, having demonstrated strong capabilities in restoring sharp boundaries in non-medical super-resolution settings. Y et, their application in 4D Flow MRI remains unexplored, with implementation challenged by known issues such as training instability and non-convergence. In this study, we investigate GAN-based super-resolution and denoising in 4D Flow MRI. Training and validation were conducted using patient-specific cerebrovascular in-silico models, converted into synthetic images via an MR-true reconstruction pipeline. A dedicated GAN architecture was implemented and evaluated across three adversarial loss functions: V anilla, Relativistic, and Wasserstein. Our results demonstrate that the proposed GAN improved near-wall velocity recovery compared to a non-adversarial reference (vNRMSE: 6.9% vs. 9.6%); however, that implementation specifics are critical for stable network training. While V anilla and Relativistic GANs proved unstable compared to generator-only training (vNRMSE: 8.1% and 7.8% vs. 7.2%), a Wasserstein GAN demonstrated optimal stability and incremental improvement (vNRMSE: 6.9% vs. 7.2%). Introduction Single image super-resolution (SISR) is the task of reconstructing a high-resolution (HR) image from its low-resolution (LR) counterpart, recovering features blurred or not fully conveyed in the input LR data. While traditional deterministic methods have been extensively studied [4, 5, 6], deep learning approaches - particularly those based on convolutional neural networks (CNNs) - have become the dominant strategy owing to their ability to learn complex spatial mappings directly from data [6, 7, 8, 9, 10, 11, 12]. In recent years, both network architectures and training strategies have evolved significantly in this domain [8, 9, 12, 13, 14, 15], consistently improving super-resolution performance.