Maximizing SNR per unit time in diffusion MRI with multiband T-Hex spirals

The characterization of tissue microstructure using diffusion MRI (dMRI) signals is rapidly evolving, with increasing sophistication of signal representations and microstructure models. However, this progress often requires signals to be acquired with very high b-values (e.g., b > 30 ms/μm² ), along many directions, and using multiple b-values, leading to long scan times and extremely low SNR in dMRI images. The purpose of this work is to boost the SNR efficiency of dMRI by combining three particularly efficient spatial encoding techniques and utilizing a high-performance gradient system (G_max  ≤ 300 mT/m) for efficient diffusion encoding.

Spiral readouts, multiband imaging, and sampling on tilted hexagonal grids (T-Hex) are combined and implemented on a 3T MRI system with ultra-strong gradients. Image reconstruction is performed through an iterative cg-SENSE algorithm incorporating static off-resonance distributions and field dynamics as measured with an NMR field camera. Additionally, T-Hex multiband is combined with a more conventional EPI-readout and compared with state-of-the-art blipped-CAIPIRINHA sampling. The advantage of the proposed approach is furthermore investigated for clinically available gradient performance and diffusion kurtosis imaging.

High fidelity in vivo images with b-values up to 40 ms/μm² are obtained. The approach provides superior SNR efficiency over other state-of-the-art multiband diffusion readout schemes.

The demonstrated gains hold promise for the widespread dissemination of advanced microstructural scans, especially in clinical populations.