Deep Learning-Based Super-Resolution Reconstruction of T2w Single-Shot Sequences in Pediatric Brain MRI

Motivation: Single-Shot sequences are essential in pediatric MRI, where motion is a major challenge, as they enable rapid image acquisition, although some image quality may be sacrificed. Balancing speed and quality is key to achieving optimal results. Goal(s): To assess the diagnostic value of Super-Resolution reconstructed T2w Single-Shot sequences in pediatric brain MRI, using an industry-developed deep-learning algorithm that combines compressed sensing with image denoising and resolution upscaling. Approach: Single-Shot sequences without (T2-SSHconv) and with Super-Resolution reconstruction (T2-SSHDL) were compared qualitatively and quantitatively. Results: T2-SSHDL not only showed improved image sharpness but also significantly increased lesion conspicuity and overall image quality. Impact: Deep learning reconstruction significantly enhances the quality of rapid T2w Single-Shot sequences in pediatric brain MRI, improving image sharpness and lesion conspicuity while maintaining motion robustness. This advancement could reduce sedation requirements in pediatric neuroimaging protocols.