English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Magnetic Resonance Imaging represents an important diagnostic modality;however, its inherently slow acquisition process poses challenges in obtainingfully sampled k-space data under motion in clinical scenarios such asabdominal, cardiac, and prostate imaging. In the absence of fully sampledacquisitions, which can serve as ground truth data, training deep learningalgorithms in a supervised manner to predict the underlying ground truth imagebecomes an impossible task. To address this limitation, self-supervised methodshave emerged as a viable alternative, leveraging available subsampled k-spacedata to train deep learning networks for MRI reconstruction. Nevertheless,these self-supervised approaches often fall short when compared to supervisedmethodologies. In this paper, we introduce JSSL (Joint Supervised andSelf-supervised Learning), a novel training approach for deep learning-basedMRI reconstruction algorithms aimed at enhancing reconstruction quality inscenarios where target dataset(s) containing fully sampled k-space measurementsare unavailable. Our proposed method operates by simultaneously training amodel in a self-supervised learning setting, using subsampled data from thetarget dataset(s), and in a supervised learning manner, utilizing data fromother datasets, referred to as proxy datasets, where fully sampled k-space datais accessible. To demonstrate the efficacy of JSSL, we utilized subsampledprostate parallel MRI measurements as the target dataset, while employing fullysampled brain and knee k-space acquisitions as proxy datasets. Our resultsshowcase a substantial improvement over conventional self-supervised trainingmethods, thereby underscoring the effectiveness of our joint approach. Weprovide a theoretical motivation for JSSL and establish a practical"rule-of-thumb" for selecting the most appropriate training approach for deepMRI reconstruction.

JSSL: Joint Supervised and Self-supervised Learning for MRI Reconstruction