Accelerating brain T2-weighted imaging using artificial intelligence-assisted compressed sensing combined with deep...

Area of Science:

• Magnetic Resonance Imaging
• Artificial Intelligence in Medical Imaging
• Neuroimaging

Background:

• T2-weighted imaging (T2WI) is crucial for detecting brain edema and lesions but suffers from long scan times and motion artifacts.
• Artificial intelligence-assisted compressed sensing (ACS) and deep learning-based reconstruction (DLR) offer solutions for accelerated MRI scans.
• The combined efficacy of ACS and DLR for T2WI at 5.0T has not been previously investigated.

Purpose of the Study:

• To evaluate the diagnostic performance of an integrated ACS-DLR technique for brain T2WI at 5.0T.
• To compare the ACS-DLR technique against conventional parallel imaging (PI) protocols.
• To assess image quality, artifacts, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR).

Main Methods:

• Prospective analysis of 98 participants undergoing brain T2WI using ACS, DLR, and PI techniques.
• Subjective evaluation of image quality and artifacts by two independent observers.
• Objective assessment of SNR and CNR in gray matter, white matter, and cerebrospinal fluid.

Main Results:

• ACS and DLR reduced acquisition time by 78%.
• Deep learning-based reconstruction (DLR) demonstrated superior overall image quality compared to ACS and equivalent quality to PI.
• DLR achieved the highest SNR and comparable CNR to PI, outperforming ACS.

Conclusions:

• The integration of ACS and DLR enables ultrafast brain T2WI acquisition at 5.0T.
• This combined technique maintains superior SNR and comparable CNR to conventional PI sequences.
• ACS-DLR represents a promising advancement for efficient and high-quality neuroimaging.