MRI has become an invaluable tool for diagnostic brain imaging, providing unrivalled qualitative and quantitative information to the radiologist. However, due to long scanning times and capital costs, access to MRI lags behind CT. Typical brain protocols, lasting between 20-50 minutes per patient, set a clear limitation to patient experience, scanner throughput and operation profitability, resulting in long waiting times for an appointment. As image quality, in terms of spatial resolution and noise, is strongly dependent on acquisition duration, significant scanning acceleration must successfully address challenging image degradation. The researchers proposed acquiring accelerated scan-pairs (e.g., with and without CMI) with orthogonal phase-encoding directions, leading to orthogonal blur directions between the individual scans. This facilitates mutual deblurring of the scan-pair, using machine learning via deep neural networks. The outcome is substantial acceleration, with no need to access the MRI internals and on top of compressed-sensing techniques. Qualitative and quantitative validation was provided against state-of-the-art deblurring methods, for an acceleration factor of 4 beyond compressed sensing own acceleration. The proposed method outperforms the compared methods, is 100% vendor agnostic, processing DICOM images in-situ or from a commercial computing cloud.

UNMET NEED
MRI scanning sessions are slow processes, typically taking 20-50 minutes per patient, limiting scanner throughput, profitability and access to this important imaging modality. Image quality is strongly dependent on acquisition duration. Therefore, significant scanning acceleration results in degraded non-diagnostic images. This research upgrades these rapidly-acquired degraded images to facilitate diagnostic use.

OUR SOLUTION
The researchers developed a neural approach that jointly deblurs scan pairs acquired with mutually orthogonal phase encoding directions. This leverages the complementarity of the respective phase encoded information as blur directions are also mutually orthogonal between the scans in the pair. The proposed architecture, trained end-to-end using machine learning via deep neural networks, is applied to High Resolution T1W scan pairs consisting of one scan with contrast media injection (CMI), and one without. Superior qualitative and quantitative comparison results were obtained against state-of-the-art deblurring methods, for an acceleration factor of 4 beyond compressed sensing acceleration, requiring less than 40 seconds to acquire a full resolution T1W brain anatomical scan. The method is vendor agnostic, processing dicom files in situ or remotely, on a commercial computing cloud. It is easily applied to any body part MRI imaging.

APPLICATIONS
Accelerated MRI scans.

STATUS
Full proof of concept, with promising experimental results.

INTELLECTUAL PROPERTY
COMPLEMENTARY PHASE ENCODING FOR PAIR-WISE NEURAL DEBLURRING OF ACCELERATED BRAIN MRI, Application/Provisional Number 63/415,015, Priority Date 11 Oct 2022
REFERENCES
G. Hod, M. Green, M. Vasserman, E. Konen, S. Shrot, I. Nelkenbaum, N. Kiryati and A. Mayer, Complementary Phase Encoding for Pair-wise Neural Deblurring of Accelerated Brain MRI, presented at ECCV 2022 Medical Computer Vision Workshop, October 2022. Available at:
https://www.researchgate.net/publication/365182136_Complementary_Phase_Encoding_for_Pair-wise_Neural_Deblurring_of_Accelerated_Brain_MRI