Fast and Sustainable Brain MRI by Triple-Modality Deblurring

Motivation: Lengthy MRI acquisition negatively impacts sustainability due to high energy consumption, increased patient discomfort, and reduced clinical efficiency. Goal(s): To develop an AI-based MRI reconstruction method that leverages complementary information from triple-modality orthogonal undersampling, significantly shortening acquisition times while maintaining image quality. Approach: We introduced orthogonal, direction-specific undersampling of k-space data (x, y, z) combined with a customized 3D slice-based DCNN for joint deblurring of accelerated T2-weighted, T2-FLAIR, and SWIp images. Results: The proposed method achieved ~4-fold acquisition acceleration compared to compressed sensing alone, with an average SSIM of 0.82, maintaining diagnostic image quality and substantially decreasing scanner operation time. Impact: The proposed method enhances MRI sustainability by significantly reducing acquisition time, energy usage (approximately 60-75% reduction in gradient and RF power), and susceptibility artifacts. Improved efficiency enables sustainable imaging practices, benefiting clinical throughput in high-volume or resource-constrained settings.