The MPRAGE sequence [1] is an integral component of clinical neuroimaging. It is also widely used for volumetry and as an anatomical reference, typically for fMRI studies. A typical 1mm³ isotropic resolution scan takes on the order of 4-5 min, making it susceptible to subject motion. We investigated a novel 3D MPRAGE sequence that employs a radial fan beam segmentation of k_y-k_z space and compared it with conventional MPRAGE in 29 patients for motion robustness and diagnostic preference.

Radial fan beam (RFB) segmentation [2] is an efficient way to traverse k-space, whilst achieving a true centric acquisition for improved contrast. 2D (k_y-k_z) k-space is segmented into M radial fan beams (Figure 1) and each fan beam is acquired following an inversion module. Within each fan beam, the N k-space points are ordered in increasing radial distance k_r to maximize image contrast. This decouples N_z, the number of slices, from N. The points are still on a Cartesian grid, enabling fast online reconstruction.

We hypothesized that RFB MPRAGE would exhibit motion robustness similar to that of 2D radial imaging, especially on pediatric and geriatric patients compared to conventional MPRAGE.

After informed consent, 29 patients were scanned on a 1.5T scanner (HDx 15M4, GE Healthcare) using an 8-channel receive array head coil. Scan parameters: 24 cm FOV, 224x224x160 acquisition matrix, 1.2 mm thick sagittal sections, 10° flip, BW +12 kHz, TR 8.5-10ms, 2X parallel imaging, 3.3 min scan time. The TS/TI of the two sequences were adjusted to maintain comparable GM-WM contrast and scan time. The two MPRAGE sequences were acquired following injection of Gadolinum contrast. In half the patients, the order of the MPRAGE sequences were swapped to remove any image quality differences arising from delayed acquisition or motion.

For image analysis, the two sequences were rated independently by two experienced neuroradiologists in a blinded fashion. Motion artifacts were ranked on a scale of 1-4: 1 – Severe, could miss lesions; 2 – Moderate, unlikely to miss lesions; 3 – Mild, does not affect image quality or lesion detection; 4 – No artifacts. Overall image preference was graded on a 0 (no preference) to 2 (strongly preferred) scale, relative to one another. These ratings were subsequently adjusted by sign to reflect preference for RFB (+) or conventional (-). Inter-observer variability was assessed using Cohen's Kappa statistic and a Wilcoxon signed rank test was used to determine statistical significance.

For motion artifacts, the mean image scores for RFB and conventional MP-RAGE were 3.82+0.39 and 2.97+0.82 for reader 1 and 3.69+0.54 and 2.86+0.88 for reader 2. Motion artifacts were significantly reduced in RFB compared to conventional MPRAGE (p < 0.05). Both readers overwhelmingly preferred RFB MPRAGE over conventional MPRAGE (mean values of 1 and 1.24 for readers 1 and 2). For both readers, RFB was preferred in 26/29 cases, conventional in 1/29 and no preference for 2/29 cases (κ = 1).

Figure 2 shows representative images from a 19-year old subject evaluated for concussion showing the effect of mild motion. RFB MPRAGE (right) is clearly superior to conventional MPRAGE (left) in minimizing artifacts and preserving image quality. Figure 3 shows representative images from a 71-year old subject who was unable to hold still. The streak artifacts in RFB MPRAGE (right) are more benign compared to the severe ghosting artifacts in conventional MPRAGE (left). Most other series in the examination were also degraded by motion artifacts. An 81-year old patient referred for glioblastoma is shown in Figure 4, again showing reduced artifacts with RFB MPRAGE (right) compared to conventional MPRAGE (left).

The radial fan beam segmentation, originally proposed for flexible choice of N (α pulses per inversion pulse) and 2D parallel imaging, was used for improving the motion robustness of MPRAGE imaging in this study. We demonstrated significantly reduced motion artifacts and improved diagnostic image quality (as measured by strong radiologist preference and scores) for the RFB MPRAGE sequence compared to conventional MPRAGE in a largely pediatric/geriatric patient population. This was achieved with no increase in scan time or advanced image or navigator-based motion correction schemes.