Referenceless 4D Flow MR for cardiac imaging

Motivation: Scan times for clinical 4D Flow MRI are still around 5-10 mins, which is too long for ideal integration in clinical practice. Goal(s): To investigate the possibility to reduce scan time by reconstructing three-directional velocity without the acquisition of reference scan in cardiac applications. Approach: Training a conditional generative adversarial network (cGAN) to estimate the reference scan from the three-velocity encoded scan segments. Results: Correlation coefficients for the pulmonary and systemic flow volumes and for maximum velocities were higher than 0.99, showing an excellent agreement between the cGAN-enabled referenceless 4D Flow MRI and conventional 4D flow MRI, thus potentially reducing scan time. Impact: By reconstructing the three-directional velocity without the reference scan it is possible reduce scan time of 4D Flow MRI, alleviating costs and patient discomfort, which especially is important for elderly and impaired subjects.