0.9mm isotropic 1min MPRAGE using highly-accelerated Deep learning Reconstruction for Brain Structural Analysis

Motivation: The project was driven by the need to reduce 3D T1-weighted MRI acquisition times, which are often prolonged, leading to motion artifacts and compromised image quality in structural nuroimaging analysis. Goal(s): To evaluate whether deep learning reconstruction can shorten MRI scan times without significantly compromising image quality, facilitating efficient clinical and research neuroimaging. Approach: We employed a deep learning technique, DL-speed, to reconstruct undersampled data from accelerated MRI scans, assessing image quality against conventional methods using a standardized rating system. Results: Images with DL-speed maintained image quality, despite a slight quality trade-off, suggesting its viability for rapid, motion-artifact-reduced neuroimaging in various patient populations. Impact: Our results impact clinicians and patients by enabling faster, high-quality MRIs, reducing patient discomfort and motion-related artifacts. This advance opens avenues for more efficient neuroimaging protocols, enhancing patient care and research productivity.