The improvement of T2 weighted and diffusion weighted image quality in breast magnetic resonance imaging by deep learning reconstruction

Motivation: Improvement of breast MRI images quality and better protocols, such as less noise, motion artifacts, better lesion characterization and less time, are still required in clinic. Goal(s): To investigate the potential benefit of deep learning reconstruction in T2 weighted and diffusion weighted images. Approach: 70 patients were performed T2 weighted, diffusion weighted images and AIR Deep Learning technology was used to do the reconstruction. Results: The image quality and visualization evaluation of deep learning (DL) reconstructed T2WI were significantly improved then T2WI, while DL-DWI didn't show advantage in compare with DWI and MUSE-DWI. Impact: This evaluation can be useful for reasonable T2WI and DWI breast imaging protocol picking based on deep learning methods, which may reducethepatients' uncomfortable or help better lesion characterization.