Primitive Representation Learning for Unsupervised Dynamic Contrast-Enhanced MRI Reconstruction

Forum topic · 小凯 · 2026-08-20

Summary

This arXiv paper (2608.18055) presents a multi-dimensional, primitive-based framework for unsupervised dynamic contrast-enhanced (DCE) MRI reconstruction. Building on prior scan-specific approaches that use Gaussian and Gabor primitives, the method extends primitive representation learning to handle the dynamic contrast dimension, which previously remained unsolved. The framework disentangles underlying anatomy, dynamic contrast enhancement, and residual motion into separate temporal basis functions, enabling a geometrical interpretation of the learned representation. According to the authors, the architecture achieves performance competitive with conventional reconstruction methods in both reconstruction quality and accuracy of extracted aortic and kidney enhancement curves, while requiring no large training datasets. Its 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 include Veronika Spieker, Wenqi Huang, Cemre Ariyurek, Daniel Rueckert, Julia A. Schnabel, and Sila Kurugol; released August 18, 2026.

Paper Overview

Field: Computer Vision / Medical Imaging Authors: Veronika Spieker, Wenqi Huang, Cemre Ariyurek, Liam Timms, Daniel Rueckert, Onur Afacan, Julia A. Schnabel, Sila Kurugol Published: 2026-08-18 arXiv: 2608.18055

Summary

Reliable quantitative analysis of dynamic contrast-enhanced (DCE) 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.

This work proposes a multi-dimensional, primitive-based framework for DCE-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.

Key Findings

The 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.
No large training datasets are required, keeping the scan-specific, unsupervised character of prior primitive-based methods.

Code

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

*Automatically collected on 2026-08-20.*

Primitive Representation Learning for Unsupervised Dynamic Contrast-Enhanced MRI Reconstruction

Summary

Paper Overview

Summary

Key Findings

Code

Tags