Evaluation of multi-shell diffusion MRI acquisition strategy on quantitative analysis using multi-compartment models.

BACKGROUND

Multi-compartment diffusion models such as Neurite Orientation Dispersion and Density Imaging (NODDI) have been increasingly used for diffusion MRI (dMRI) data processing in biomedical research. However, those models usually require multiple HARDI shells that may increase scanning duration substantially, and their application can be hindered in uncooperative patients (like infants) accordingly. Also, it is highly expected that the same dataset can be explored with multiple diffusion models for retrieving complementary information.

METHODS

Multiple gradient-encoding schemes which consisted of 4-6 shells, moderate b-values (bmax =1,500 or 2,000 s/mm), and 32-80 gradient directions were explored. The corresponding time of acquisition (TA) for a single scan ranged from 3 to 8 minutes respectively. The dMRI protocols were tested on macaque monkeys using a 3T clinical setting. The data were analysed using both NODDI and diffusion basic spectrum imaging (DBSI) models.

RESULTS

The maps of orientation dispersion index (ODI) and CSF were consistent across the 4-6 shell sampling schemes. However, the corresponding intra-cellular volume fraction (ICVF) maps showed reduced pixel counts [1,100±98 (80 directions) 806±70 (32 directions), one slice] in white matter when fewer gradient directions or lower b-value was applied. The hindered diffusion and CSF ratio maps were comparable across these sampling schemes. The maps of restricted diffusion ratio varied across the schemes. However, its mean ratios (0.23±0.02 0.22±0.01) and pixel counts (1,540±70 1,510±38, one slice) between the schemes of 80 and 32 directions with b=2,000 s/mm were comparable.

CONCLUSIONS

The present study reports a fast multi-shell dMRI data acquisition and processing strategy which allows for obtaining complementary information about microstructural alteration and inflammation from a single dMRI data set with both NODDI and DBSI models. The proposed approach may be particularly useful for characterizing the neurodegenerative disorders in uncooperative patients like children or acute stroke patients in which brain injury is associated with inflammation.