Deep learning denoising reconstruction enables faster T2-weighted FLAIR sequence acquisition with satisfactory image...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Deep learning reconstruction (DLR) aims to reduce MRI scan times without sacrificing image quality.
• The clinical effectiveness of DLR for 2D T2-weighted FLAIR brain imaging requires further evaluation.

Purpose of the Study:

• To assess a commercial DLR technique for accelerated 2D T2-weighted FLAIR brain MRI.
• To determine if DLR reduces scan time while maintaining image quality and diagnostic accuracy.

Main Methods:

• 47 participants underwent standard-of-care (SOC) and accelerated DLR T2-weighted FLAIR MRI.
• Two readers subjectively assessed image quality, lesion conspicuity, SNR, CNR, and artifacts.

Main Results:

• A strong preference for SOC FLAIR was noted for overall image quality and direct comparison.
• No significant differences were found in lesion conspicuity, SNR, or CNR between sequences.
• DLR images showed significantly more phase ghosting and pseudolesions.

Conclusions:

• DLR enables faster FLAIR acquisition with comparable image quality and lesion conspicuity.
• Increased artifacts (phase ghosting, pseudolesions) with DLR may negatively impact reading speed and diagnostic confidence.