Figure 5.

Images obtained in a patient with mild carotid stenosis. Maximum intensity projection (MIP) images comparing ungated QISS MRA in a patient with: (a) 1-shot Cartesian k-space sampling (2 min 21 s); (b) 1-shot radial, (c) 2-shot radial and (d) 3-shot radial k-space sampling (scan times of 2 min 3 sec, 4 min 42 s, 7 min 3 s, respectively); and (e) contrast-enhanced MRA (CEMRA) obtained 9 weeks prior to QISS MRA. Arrow in (a) points to a mild stenosis of the left carotid bifurcation. Note the reduced pulsation artifact and improved image quality obtained with 1-shot radial sampling as compared with 1-shot Cartesian sampling, the similar appearances of the radial QISS acquisitions despite major differences in scan time, and the correlation of radial QISS with CEMRA. Panels (f) and (g): Sagittal 2-mm-thick multiplanar reformations through the proximal internal carotid artery showing comparisons of standard and deep learning processed QISS reconstructions obtained with (f) 1-shot and (g) 3-shot radial sampling. Deep learning processing reduced signal variability in background tissue (false color insets) without degrading small arterial detail (dashed arrows).