The orientationally-averaged diffusion magnetic resonance (MR) signal acquired at high diffusion weighting shows great potential for answering fundamental questions about neural tissue microstructure[1]. The noise-induced bias in the magnitude-valued signal and angular resolution limitations in diffusion encoding are among the challenges in obtaining an accurate estimate. Here, we present a data processing framework for computing the orientationally-averaged diffusion signal that corrects the noise induced bias and accounts for the low angular resolution of the acquisition. Noise correction is performed using a statistical transformation framework [2] that converts the noisy MR signal from a noncentralChi distribution to a noisy Gaussian one. Weights for each of the probing directions are computed to improve the rotationally invariant representation of the sample. Synthetic data, generated to mimic diffusion acquisitions with different noise levels and number of acquisition directions, were used to test the data processing framework. The performance of the framework was evaluated by comparing the processed data with the analytical solution of the orientationally-averaged signal. Results show that the computation of the orientationally-averaged signal benefits from both the noise correction and the weighted averaging, especially in the low signal regime. This work provides a tool for processing high diffusion-weighted MR signals whose interpretation could improve our knowledge about neural tissue microstructure.

[1] Özarslan E, Yolcu C, Herberthson M, Knutsson H, Westin CF. Influence of the size and curvedness ofneural projections on the orientationally averaged diffusion MR signal. Front Phys, 2018; 6:17.

[2] Koay CG, Özarslan E, Basser PJ. A signal transformational framework for breaking the noise floorand its applications in MRI. J Magn Reson 2009; 197(2):108–119.