Primitive Representation Learning for Unsupervised Dynamic Contrast-Enhanced MRI Reconstruction (arXiv 2608.18055)

Forum topic · 小凯 · 2026-08-20

Summary

This post introduces an arXiv paper (2608.18055) proposing a multi-dimensional, primitive-based unsupervised framework for dynamic contrast-enhanced (DCE) MRI reconstruction. Building on scan-specific reconstructions that use Gaussian and Gabor primitives—effective without large training datasets—the method extends to the additional temporal dimension of dynamic contrast. It disentangles underlying anatomy, dynamic contrast enhancement, and residual motion into separate temporal basis functions, enabling a geometrical interpretation of the learned representation. Experiments show performance competitive with conventional reconstruction methods in both image reconstruction quality and accuracy of extracted aortic and renal enhancement curves. A modular hierarchical design naturally extends to additional dynamic factors and higher acceleration rates. Code is available at https://github.com/compai-lab/2026-GaborDCE-spieker. Authors: Veronika Spieker, Wenqi Huang, Cemre Ariyurek, Liam Timms, Daniel Rueckert, Onur Afacan, Julia A. Schnabel, Sila Kurugol.

Paper Overview

Field: Computer Vision / Medical Imaging
arXiv: 2608.18055
Authors: Veronika Spieker, Wenqi Huang, Cemre Ariyurek, Liam Timms, Daniel Rueckert, Onur Afacan, Julia A. Schnabel, Sila Kurugol

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. The authors 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. This architecture achieves performance competitive with conventional reconstruction methods, both in reconstruction quality and in the accuracy of extracted aortic and kidney enhancement curves. The modular hierarchical design naturally extends to additional dynamic factors and higher acceleration rates.

Key Points

Extends primitive-based (Gaussian/Gabor) scan-specific MRI reconstruction to dynamic contrast-enhanced imaging without requiring large training datasets.
Disentangles anatomy, contrast enhancement, and residual motion into separate temporal basis functions.
Enables geometric interpretability of the learned representation.
Matches conventional methods in reconstruction quality and pharmacokinetic curve extraction (aorta and kidneys).
Modular design scales to more dynamic factors and higher undersampling rates.

Code

https://github.com/compai-lab/2026-GaborDCE-spieker