Deep Learning-Based Reconstruction of 3D T1 SPACE Vessel Wall Imaging Provides Improved Image Quality with Reduced...

Area of Science:

• Neuroimaging
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Intracranial vessel wall imaging presents technical challenges, requiring high spatial resolution, effective signal suppression, and short scan times.
• Current methods often struggle to balance these competing demands for optimal clinical application.

Purpose of the Study:

• To evaluate a novel deep learning-optimized T1-weighted sequence for intracranial vessel wall imaging.
• To compare the performance of deep learning-based reconstruction against conventional methods.

Main Methods:

• T1 3D Sampling Perfection with Application optimized Contrast using different flip angle Evolution (SPACE) sequences were assessed.
• Deep learning-based image reconstruction was compared with clinical sequences in healthy controls and patients.
• Neuroradiologists rated vessel wall/lumen delineation, background noise, sharpness, and CSF signal on a Likert-like scale.

Main Results:

• Deep learning reduced scan time from 7:26 to 5:23 minutes.
• Significantly improved scores for vessel wall signal and lumen visualization were observed with deep learning.
• Deep learning reconstruction demonstrated lower noise, enhanced sharpness, and uniform CSF signal.

Conclusions:

• Deep learning-optimized sequences offer shorter gradient times and superior intracranial vessel wall visualization.
• These improvements in image quality may facilitate broader clinical implementation of vessel wall imaging.
• Further validation in larger patient cohorts is warranted.