Proximal-Nets: Unfolding Proximal Algorithms for Accelerating and Improving MR Image Reconstruction

Researchers: Kaixuan Wei, Angelica Aviles-Rivero, Carola Schönlieb

Magnetic Resonance Imaging (MRI) is a non-invasive imaging technique providing both functional and anatomical information for clinical diagnosis. Imaging speed is a fundamental challenge. Fast MRI techniques are essentially demanded for accelerating data acquisition while still reconstructing high-quality image. Therefore, a central limitation of Magnetic Resonance Imaging (MRI) is the linear relation between the number of necessary measurements to form an image and the acquisition time. This constraint causes negative effects including: (i) sensitivity to motion causing image degradation, (ii) reduced clinical throughput and (iii) patient non-compliance generating more artefacts during the image formation. Thus, a central challenge in MRI is to respond to the question − How to decrease the long data-acquisition time? To this aim, our research question will be oriented to the Compressed Sensing (CS) setting.

A central topic in CS-MRI is how to choose an optimal image transform domain/subspace and the corresponding sparse regularisation. With this purpose in mind, we aim to design a fast yet accurate method to reconstruct high-quality MR images from under-sampled k-space data, this, by combining CS implications and deep learning.

Who's involved