Deep Learning Reconstruction for Accelerating Synthetic MRI in Breast Imaging: A Comparative Analysis of Accelerated and Standard Protocols

Motivation: The scanning time for Synthetic MRI (SyMRI) remains relatively long, which limits its widespread clinical use. Addressing the need to reduce scanning time while ensuring image quality is crucial. Goal(s): To explore the potential of deep learning-based reconstruction(DLR) in accelerating SyMRI while maintaining stable quantitative parameter values and image quality. Approach: A total of 58 female patients were included. 3x-SyMRI-DLR, 3x-SyMRI, 2x-SyMRI images were compared for quantitative parameters (T1, T2, PD), image signal-to-noise ratio (SNR), subjective image quality, and diagnostic performance. Results: The 3x-SyMRI-DLR protocol significantly reduces scanning time while maintaining stable T1, T2, and PD quantitative measurements. Impact: Deep learning reconstruction for accelerating synthetic MRI is poised to enhance the clinical application of synthetic MRI in diagnosing breast diseases, thereby improving examination efficiency.