Sodium (²³Na) ultra-short echo time imaging in the human brain using a 3D-Cones trajectory

Object: Sodium magnetic resonance imaging (²³Na-MRI) of the brain has shown changes in ²³Na signal as a hallmark of various neurological diseases such as stroke, Alzheimer's disease, Multiple Sclerosis and Huntington's disease. To improve scan times and image quality, we have implemented the 3D-Cones (CN) sequence for in vivo ²³Na brain MRI. Materials and methods: Using signal-to-noise (SNR) as a measurement of sequence performance, CN is compared against more established 3D-radial k-space sampling schemes featuring cylindrical stack-of-stars (SOS) and 3D-spokes kooshball (KB) trajectories, on five healthy volunteers in a clinical setting. Resolution was evaluated by simulating the point-spread-functions (PSFs) and experimental measures on a phantom. Results: All sequences were shown to have a similar SNR arbitrary units (AU) of 6-6.5 in brain white matter, 7-9 in gray matter and 17-18 AU in cerebrospinal fluid. SNR between white and gray matter were significantly different for KB and CN (p = 0.046 and <0.001 respectively), but not for SOS (p = 0.1). Group mean standard deviations were significantly smaller for CN (p = 0.016). Theoretical full-width at half-maximum linewidth of the PSF for CN is broadened by only 0.1, compared to 0.3 and 0.8 pixels for SOS and KB respectively. Actual image resolution is estimated as 8, 9 and 6.3 mm for SOS, KB and CN respectively. Conclusion: The CN sequence provides stronger tissue contrast than both SOS and KB, with more reproducible SNR measurements compared to KB. For CN, a higher true resolution in the same amount of time with no significant trade-off in SNR is achieved. CN is therefore more suitable for ²³Na-MRI in the brain.