Super-Resolution Image Enhancement for 3D Morphological Lung MRI with bSTAR at 0.55T

Motivation: Despite the development of various dedicated techniques for morphological lung MRI, these methods still offer limited spatial resolution compared to computed tomography. Goal(s): With the rapid progress in AI-based image enhancement techniques, our study aims to improve lung MRI quality using a deep learning-based super-resolution model, thereby enabling better visualization of lung morphology. Approach: We applied a pre-trained 2D image enhancement model to 3D lung MRI datasets acquired at 0.55 T to enhance image quality. Results: The multi-orientation technique we implemented preserves anatomical accuracy while reducing noise, leading to significant improvements in lung MRI image quality. Impact: Due to physical and physiological challenges lung MRI is currently rarely used in clinical routine1. Our study demonstrates that a deep learning-based super-resolution image enhancement results in improved visualization of MRI lung morphology without introducing new anatomical structures.