Synopsis

In this study, we measure the reproducibility of metrics derived from the ensemble averaged propagator (EAP) estimated using the 3D-SHORE technique using a clinically feasible high angular resolution diffusion dataset repeated 11 times on a single subject. We find very low reproducibility in measures of microstructural restriction (including the mean-squared displacement and return to origin probability) as well as measures of the orientation distribution function (including peaks of the ODF). This study highlights the limitations of using advanced models with empirical data, particularly in the low-SNR regime.

PURPOSE:

METHODS:

Acquisition: One healthy subject was imaged on a 3T Phillips Achieva Scanner in three different scan sessions on three consecutive days. The scanning protocol consisted of five diffusion shells (b-values of 1000, 1500, 2000, 2500, and 3000 s/mm2) with 96 DW directions each. The data was acquired at 2.5mm isotropic resolution, with a matrix of 96 x 96 x 38 slices, using a DW PGSE-EPI sequence. This protocol was repeated four times on days 1 and 2, and three times on day 3, resulting in a total of 11 repeats of each diffusion shell. For each set of 5 shells, an additional “b0” was acquired with reverse phase encoded volumes to be used for susceptibility distortion and eddy current correction.

EAP Processing: We model the diffusion signal using the 3D Simple Harmonic Oscillator based Reconstruction and Estimation (3D SHORE) [2, 3]. Orientation features of the EAP are extracted, including the ODF and local maxima of the ODF, and microstructure related indices are extracted including the return to origin probability (RTO) and mean squared displacement (MSD). These indices reflect the degree of restriction of water molecules and have been directly linked to compartment size and shapes [3, 4].

Reproducibility: The voxel-wise mean and standard deviation of RTO and MSD are calculated across all 11 repeat scans, followed by calculation of the coefficient of variation (CV). To assess reproducibility of the ODF, the Jensen-Shannon Divergence (JSD) is calculated (as in [5]), which measures similarity of probability distributions. Finally, the angular variation of the primary (AV1) and secondary (AV2) peaks of the ODF are computed [6]. In addition to voxel-wise metrics, the average value within 5 ROIs are calculated. White matter ROI include the splenium of the corpus callosum (SCC), the internal capsule (IC), and the centrum semiovale (CS). Gray matter (GM) structures include the sub-cortical globus pallidus (GP) in addition to the precentral gyrus (PCG) of the cerebral cortex.

RESULTS:

Figure 1 shows reproducibility results for the measures MSD and RTO across 11 repeats. Maps of mean and CV of the measures are shown on top, in addition to ROI-averaged mean, standard deviation, and CV in table form below. The CV for both measures is >10% in all cases (indicating a low reproducibility), with particularly high CV in the SCC and GP.

Figure 2 shows maps of JSD, AV1, and AV2 in a selected axial slice, in addition to the ROI-averaged metrics in table form. Again, the JSD is particularly high in the SCC and GP, and lower in the cortex. The AV1 is low in most WM regions (except the CS which contains multiple fiber populations), while the AV2 is consistently larger than 25°.

DISCUSSION:

CONCLUSION:

Acknowledgements

References