Fast Submillimeter Whole-Brain T₂*-Weighted Imaging Using 3D-EPI With CAIPIRINHA and Deep-Learning Denoising at 3T

Area of Science:

• Neuroimaging
• Medical Physics
• Radiology

Background:

• Multiple sclerosis (MS) diagnosis requires high-resolution T₂*-weighted (T₂*w) imaging for novel biomarkers.
• Conventional T₂*w 3D gradient-echo sequences are too slow for clinical use at submillimeter resolution.
• Echo planar imaging (EPI) offers faster acquisition but traditionally has lower resolution and quality.

Purpose of the Study:

• To evaluate a segmented 3D echo planar imaging (3D-EPI) sequence for fast, submillimeter whole-brain T₂*w MRI.
• To assess the impact of Controlled Aliasing in Parallel Imaging Results in Higher Acceleration (CAIPIRINHA) and deep learning denoising on image quality and scan time.
• To determine the feasibility of this accelerated 3D-EPI approach for MS diagnostics.

Main Methods:

• Fifty-two participants underwent 3T MRI using a 3D-EPI sequence with CAIPIRINHA acceleration (R=2, 3, 4).
• Images were denoised using a deep learning-based convolutional neural network (DnCNN).
• Quantitative metrics (PSNR, SSIM, contrast) and blinded radiologist assessment evaluated image quality.

Main Results:

• Fast submillimeter (0.65 mm) whole-brain T₂*w imaging was achieved with scan times as low as 1 min 56 s (R=4).
• Deep learning denoising significantly improved PSNR (+3 dB), SSIM (+13%), and lesion-to-vein contrast (+9%) even at R=4.
• Accelerated 3D-EPI with denoising maintained diagnostic image quality suitable for detecting biomarkers.

Conclusions:

• The combination of 3D-EPI, CAIPIRINHA acceleration, and deep learning denoising enables rapid submillimeter whole-brain T₂*w MRI at 3T.
• This technique overcomes the limitations of conventional sequences, offering a clinically viable solution for MS diagnosis.
• The method demonstrates potential for efficient detection of MS-specific imaging biomarkers.