Abstract Background While using AI-assisted compressed sensing (ACS) combined with deep-learning reconstruction (DR) techniques has the potential to Shorten the acquisition time of brain T2-weighted imaging (T2WI), its imaging improvements and clinical application in brain imaging remains under explored. Purpose To evaluate the clinical feasibility of DR-ACS for brain T2-weighted imaging (T2WI), focusing on image quality, acquisition efficiency, and diagnostic accuracy, compared with the routine T2WI. Material and Methods A prospective cohort of 110 participants underwent brain MRI using three protocols at a 3.0-T MR scanner: routine T2WI, ACS-T2WI (without DR), and DR-ACS-T2WI. Subjective image quality (overall image quality, motion artifact, and diagnostic confidence, with a 5-point scale) and objective metrics (Signal-to-Noise Ratio (SNR), Contrast-to-Noise Ratio (CNR), and scan time) were compared. Statistical analysis included ANOVA and kappa. Results The overall image quality, motion artifact, and diagnostic confidence scores of DR-ACS-T2WI, assessed by two radiologists, were 4.90±0.30, 4.91±0.29 , 4.92±0.28(Reader 1) and 4.90±0.30 , 4.91±0.28, 4.91±0.30 (Reader 2), higher than those of ACS-T2WI and routine T2WI. DR-ACS-T2WI demonstrated superior SNRs in both white matter (65.06±12.1) and gray matter (97.25±18.52) compared to ACS-T2WI (47.62±8.65 and 71.54±12.05, respectively) and routine T2WI (34.32±6.51 and 51.92±8.62, respectively; P Conclusion The ACS combined with DR is clinically feasible for MRI examinations of brain diseases, offering significantly shorter image acquisition time and higher image quality compared with the routine T2WI.
Lai et al. (2025) studied this question.