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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Magnetic resonance imaging (MRI) is an essential diagnostic tool that suffersfrom prolonged scan times. Reconstruction methods can alleviate this limitationby recovering clinically usable images from accelerated acquisitions. Inparticular, learning-based methods promise performance leaps by employing deepneural networks as data-driven priors. A powerful approach uses scan-specific(SS) priors that leverage information regarding the underlying physical signalmodel for reconstruction. SS priors are learned on each individual test scanwithout the need for a training dataset, albeit they suffer fromcomputationally burdening inference with nonlinear networks. An alternativeapproach uses scan-general (SG) priors that instead leverage informationregarding the latent features of MRI images for reconstruction. SG priors arefrozen at test time for efficiency, albeit they require learning from a largetraining dataset. Here, we introduce a novel parallel-stream fusion model(PSFNet) that synergistically fuses SS and SG priors for performant MRIreconstruction in low-data regimes, while maintaining competitive inferencetimes to SG methods. PSFNet implements its SG prior based on a nonlinearnetwork, yet it forms its SS prior based on a linear network to maintainefficiency. A pervasive framework for combining multiple priors in MRIreconstruction is algorithmic unrolling that uses serially alternatedprojections, causing error propagation under low-data regimes. To alleviateerror propagation, PSFNet combines its SS and SG priors via a novelparallel-stream architecture with learnable fusion parameters. Demonstrationsare performed on multi-coil brain MRI for varying amounts of training data.PSFNet outperforms SG methods in low-data regimes, and surpasses SS methodswith few tens of training samples.

Learning Deep MRI Reconstruction Models from Scratch in Low-Data Regimes