aiqbench analysis
This research introduces a primitive-based framework using Gaussian and Gabor functions to reconstruct dynamic contrast-enhanced MRI scans without relying on large training datasets. By disentangling anatomy, contrast dynamics, and motion into separate temporal basis functions, the method offers a geometrical interpretation that matches conventional reconstruction quality even at high acceleration rates. The approach advances unsupervised learning in medical imaging by reducing data dependency while maintaining accuracy for clinical metrics like enhancement curves.
Primitive Representation Learning for Unsupervised Dynamic Contrast Enhanced MRI Reconstruction
arXiv:2608.18055v2 Announce Type: replace-cross Abstract: Reliable quantitative analysis of dynamic contrast-enhanced MRI requires high-quality spatiotemporal reconstructions at high undersampling rates. Scan-specific reconstructions using Gaussian and Gabor primitives have shown promising results without the need for large training datasets, but have not addressed the additional dimension of dynamic contrast. We propose a multi-dimensional, primitive based framework for dynamic contrast-enhanced MRI reconstruction that disentangles the underlying anatomy, the dynamic contrast enhancement, and residual motion into separate temporal basis functions, thereby enabling a geometrical interpretation of the representation. We show that this architecture achieves performance competitive with conventional reconstruction methods, both in reconstruction quality and in the accuracy of extracted aorta and kidney enhancement curves. The modular tier design extends naturally to additional dynamic factors and higher acceleration rates. Code available at https://github.com/compai-lab/2026-GaborDCE-spieker.
Abstract:Reliable quantitative analysis of dynamic contrast-enhanced MRI requires high-quality spatiotemporal reconstructions at high undersampling rates. Scan-specific reconstructions using Gaussian and Gabor primitives have shown promising results without the need for large training datasets, but have not addressed the additional dimension of dynamic contrast. We propose a multi-dimensional, primitive based framework for dynamic contrast-enhanced MRI reconstruction that disentangles the underlying anatomy, the dynamic contrast enhancement, and residual motion into separate temporal basis functions, thereby enabling a geometrical interpretation of the representation. We show that this architecture achieves performance competitive with conventional reconstruction methods, both in reconstruction quality and in the accuracy of extracted aorta and kidney enhancement curves. The modular tier design extends naturally to additional dynamic factors and higher acceleration rates. Code available at this https URL.
Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG); Signal Processing (eess.SP); Medical Physics (physics.med-ph)
arXiv:2608.18055 [eess.IV]
(or arXiv:2608.18055v2 [eess.IV] for this version)
Full text under the publisher's open licence · 155 words
Related coverage
PIRAMID: Progress and Plans
LessWrong · Jul 27, 2026
EndoWave: 4D Gaussian Splatting with Rational Wavelet for Endoscopic Reconstruction
arXiv · Oct 6, 2026
Unsupervised Anomaly Detection for Image Dataset Quality Assurance in Multi-Center Breast MRI
arXiv · Aug 18, 2026
Atrial Fibrillation Detection with Arbitrary Leads via a Codebook-Based Reconstruction-Classification Framework
arXiv · Aug 20, 2026