Deep Learning Accelerated Single Breath-hold abdominal DWI with full liver and kidney coverage

Motivation: Free-breathing (FB) abdominal DWI is prone to motion artifacts, leading to unstable and unreliable ADC map estimations. Goal(s): Implement and evaluate a single breath-hold DWI sequence utilizing deep learning (DL) reconstruction. Approach: Healthy volunteer underwent FB DWI, respiratory triggered (RT) DWI and single breath-hold DWI accelerated by deep learning reconstruction. Image quality and ADC histograms were compared. Results: DL-reconstructed single breath-hold DWI showed more reliable and stable ADC estimations, with fewer motion artifacts, image mismatches, and slice discontinuities compared to FB DWI and RT DWI. Impact: Single breath-hold DWI can shorten scanning time, reduce artifacts and improve ADC estimation, potentially improving diagnostic reliability in abdominal MRI.