Effects of signal-to-noise ratio on the accuracy and reproducibility of diffusion tensor imaging-derived fractional anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5 T

Abstract

Purpose: To develop an experimental protocol to calculate the precision and accuracy of fractional anisotropy (FA), mean diffusivity (MD), and the orientation of the principal eigenvector (PEV) as a function of the signal-to-noise ratio (SNR) in vivo.

Materials and methods: A healthy male volunteer was scanned in three separate scanning sessions, yielding a total of 45 diffusion tensor imaging (DTI) scans. To provide FA, MD, and PEV as a function of SNR, sequential scans from a scan session were grouped into nonintersecting sets. Analysis of the accuracy and precision of the DTI-derived contrasts was done in both a voxel-wise and region of interest (ROI)-based manner.

Results: An upward bias of FA and no significant bias in MD were present as SNR decreased, confirming results from simulation-based studies. Notably, while the precision of the PEV became worse at low SNR, no bias in the PEV orientation was observed. Overall, an accurate and precise quantification of FA values in GM requires substantially more SNR than the quantification of white matter (WM) FA values

Conclusion: This study provides guidance for FA, MD, and PEV quantification and a means to investigate the minimal detectable differences within and across scan sessions as a function of SNR.

Figure 1

Images showing the relationship between the accuracy and precision of FA and the number of DTI scans used in the diffusion tensor calculation. Results are shown at two slice levels. (a) Representative FA maps computed from 1, 3 and 15 DTI scans, respectively. (b) Difference between the average of 15 FA maps calculated from single DTI scans and the gold standard (left). Difference between the average of 5 FA maps calculated from sets of three DTI scans and the gold standard (right). (c) Standard deviation and (d) coefficient of variation of FA calculated from the 15 observations of FA maps from single DTI scans (left) and the 5 observations of FA maps calculated from sets of 3 DTI scans (right).

Figure 2

The standard deviation of voxel-wise FA measures in WM (FA_gs > 0.25) displayed as a function of fiber orientation. (a) σ(FA) computed from single DTI scans and from sets of 3 DTI scans, top and bottom surfaces respectively. (b) Results from (a) represented on the surface of a sphere with the DW directions (solid markers). The patient coordinate system is abbreviated as inferior superior (IS), anterior posterior (AP) and right left (RL).

Figure 3

Locations of regions of interest (ROIs) at four slice levels used for ROI-based analyses. These include internal capsule (ic), frontal white matter (fw), centrum semiovale (cs), globus pallidus (gp), putamen (put) and splenium of the corpus callosum (scc).

Figure 4

ROI-based FA and MD as a function of the number of scans. (a) The mean FA in an ROI and (c) the mean MD in an ROI for each unique grouping of DTI data (solid circles). The 3 scan sessions are represented by red, green and blue solid circles. (b and d) All data points from the three scan sessions in Figure 4a and 4c are consolidated to representative single values in Figure 4b and 4d, respectively (solid circles). The spread of the FA measurements in Figure 4a and the spread of MD measurements in Figure 4c were computed as the standard deviation of all points over all sessions (shown as dotted lines). The standard deviation of FA measurements within each ROI in Figure 4a was computed and averaged over all observations in all sessions (shown as error bars). Figure 4d shows the corresponding results for MD. Anatomical locations and abbreviations are shown in Figure 3.

Figure 5

ROI-based eigenvalues (λ₁, λ₂ and λ₃) as a function of the number of scans. (a) The mean eigenvalue in the putamen and (c) the mean eigenvalue in the internal capsule, for each unique grouping of DTI data. The 3 scan sessions are represented by red, green and blue solid circles. (b and d) All data points from the three scan sessions in Figure 5a and 5c are consolidated to representative single values in Figure 5b and 5d, respectively (solid circles). The spread of the eigenvalue measurements in Figure 5a and 5c were computed as the standard deviation of all points over all sessions (shown as dotted lines). The standard deviation of the eigenvalue measurements in the putamen ROI in Figure 5a was computed and averaged over all observations in all sessions (shown as error bars). Figure 5d shows the corresponding results for the internal capsule ROI. Anatomical locations and abbreviations are shown in Figure 3.

Figure 6

Reproducibility of voxel-wise FA and MD measurements. Intra-session reproducibility: (a) coefficient of variation of FA (b) coefficient of variation of MD. Inter-session reproducibility: (c) test retest variation of FA (d) test retest variation of MD. Anatomical locations and abbreviations are shown in Figure 3.

Figure 7

Reproducibility of ROI-based FA and MD measurements. Intra-session reproducibility: (a) coefficient of variation of FA (b) coefficient of variation of MD. Inter-session reproducibility: (c) test retest variation of FA (d) test retest variation of MD. Anatomical locations and abbreviations are shown in Figure 3.

Figure 8

Improvements in color-coded PEV orientation maps as a function of the number of scans. (a) Representative PEV orientation maps computed from 1, 3 and 15 DTI scans, at two slice levels. The precision and accuracy of the PEV orientation within a scan session were computed on a voxel by voxel basis in 4 anatomical regions. (b) Angular variation (AV) and (c) angular bias (AB) of the PEV as a function of SNR. Anatomical locations and abbreviations are shown in Figure 3.

Figure 9

Reproducibility of the PEV orientation across scan sessions. (a) The spatial distribution of the mean angular difference (MAD) of PEV maps computed using sets of 1, 3, and 15 scans (b) MAD as a function of the number of scans for 4 anatomical regions. Anatomical locations and abbreviations are shown in Figure 3.