[Paper Review] Gabor Primitives for Accelerated Cardiac Cine MRI Reconstruction

Wenqi Huang, Veronika Spieker+7 more|arXiv (Cornell University)|Mar 5, 2026

Advanced MRI Techniques and Applications0 citations

TL;DR

The paper introduces complex-valued Gabor primitives for MRI reconstruction to efficiently represent high-frequency content, enabling scan-specific, motion-aware cardiac cine reconstruction with a two-component low-rank temporal model.

ABSTRACT

Accelerated cardiac cine MRI requires reconstructing spatiotemporal images from highly undersampled k-space data. Implicit neural representations (INRs) enable scan-specific reconstruction without large training datasets, but encode content implicitly in network weights without physically interpretable parameters. Gaussian primitives provide an explicit and geometrically interpretable alternative, but their spectra are confined near the k-space origin, limiting high-frequency representation. We propose Gabor primitives for MRI reconstruction, modulating each Gaussian envelope with a complex exponential to place its spectral support at an arbitrary k-space location, enabling efficient representation of both smooth structures and sharp boundaries. To exploit spatiotemporal redundancy in cardiac cine, we decompose per-primitive temporal variation into a low-rank geometry basis capturing cardiac motion and a signal-intensity basis modeling contrast changes. Experiments on cardiac cine data with Cartesian and radial trajectories show that Gabor primitives consistently outperform compressed sensing, Gaussian primitives, and hash-grid INR baselines, while providing a compact, continuous-resolution representation with physically meaningful parameters.

Motivation & Objective

• Motivation to accelerate cardiac cine MRI under high undersampling
• Develop an explicit, interpretable primitive basis for MRI reconstruction
• Exploit spatiotemporal redundancy with a low-rank temporal model
• Demonstrate improvements over CS, Gaussian primitives, and hash-grid INRs on Cartesian and radial trajectories

Proposed method

• Define complex-valued Gabor primitives that place spectral support at arbitrary k-space locations via complex exponential modulation
• Model the image as a sum of N Gabor primitives with per-frame geometry and weight parameters
• Introduce a two-component low-rank temporal model: a geometry basis for cardiac motion and an intensity basis for contrast changes
• Couple geometry and weight dynamics through per-primitive temporal coefficients and low-rank bases
• Formulate a multi-coil forward model with a forward operator on the undersampled trajectory and a composite loss combining data fidelity, weight sparsity, and temporal regularization

Experimental results

Research questions

• RQ1 Does spectrally shifted (Gabor) primitives improve high-frequency content representation in undersampled cardiac MRI compared to Gaussian primitives?
• RQ2 Can a two-component low-rank temporal model capture geometry (motion) and intensity (contrast) changes efficiently in cardiac cine MRI?
• RQ3 Do Gabor primitives outperform conventional CS, Gaussian primitives, and Hash-INR baselines across Cartesian and radial undersampling scenarios?
• RQ4 Is the resulting representation compact and interpretable while maintaining or improving reconstruction quality?

Key findings

• Gabor primitives achieve the highest PSNR and SSIM across Cartesian and radial undersampling in the provided datasets
• Gabor outperforms Gaussian primitives and CS baselines, with notable gains on radial data (≈2.34 dB PSNR over PICS)
• The representation is compact (rho < 0.5) with a physically interpretable spectral distribution across k-space
• Gabor-based reconstructions show improved boundary sharpness and reduced background noise in qualitative comparisons
• The method provides a continuous, super-resolution capable representation without retraining
• Hash-INR uses more parameters and still underperforms compared to Gabor primitives in this setting