Feasibility of deep learning k-space-to-image reconstruction for diffusion weighted imaging in patients with breast cancers: Focus on image quality and reduced scan time

Purpose: This study aimed to evaluate the feasibility of accelerated DLR (deep learning reconstruction) single-shot echo planar imaging (ss-EPI) for diffusion-weighted image (DWI) in patients with breast cancers in comparison to conventional ss-EPI.Methods: Between August 2021 and February 2022, eighty-seven patients with pathologically proven breast cancer underwent DCE breast MRI including ss-EPI and DLR ss-EPI DWI sequences (TA, 3:36 min and 1:54 min, respectively) at 3 Tesla. In a randomized and blinded manner, two radiologists independently performed qual-itative analyses for overall image quality using a 5-point scale of the following components: homogeneous fat suppression, image blurring, artifact, and lesion conspicuity. Quantitative analyses were performed by mea-surement of ADC values, SNR, CNR, and lesion contrast.Results: DLR ss-EPI showed better image quality scores, CNR, and lesion contrast than ss-EPI (all P < 0.05) while reducing scan time by 47.2 %. DLR ss-EPI showed no significant difference in SNR and tumor ADC values compared to -ss-EPI (P = 0.307 and P = 0.123, respectively).Conclusions: DLR ss-EPI showed better results in the qualitative and quantitative analysis than conventional ss-EPI despite reducing scan time by 47.2%.