Skip to main content
. 2013 Jan 15;65:433–448. doi: 10.1016/j.neuroimage.2012.10.022

Fig. 12.

Fig. 12

Crossing-fibre volume fraction resolution results: percentage of successful trials (vertical axis) vs. relative voxel contributions (horizontal axis, ticks denotes the percentage of the signal provided by the axially aligned fibre). (A) CSHD volume fractions resolution at SNR = 10 vs. a T = 0.9 target. Note that as Lmax increases at this low SNR, the performance of the C = 0.9 (i.e. C = T) calibration deteriorates while C = 0.6 and C = 0.7 calibrations slightly improve. (B) CSD volume fraction resolution at SNR = 30 vs. a T = 0.6 target. Note that while increased Lmax can improve best-case volume-fraction resolution (follow the C = 0.6 calibration), it does so at the cost of an increased sensitivity to miscalibration (follow C = 0.7 and C = 0.9). (C) CSD volume fraction resolution at SNR = 50 vs. a T = 0.3 target. CSHD seems more able to resolve volume fractions at low anisotropy than dRL in comparable situations (subsection F). (D) dRL volume fraction resolution at SNR = 10 vs. a T = 0.9 target. Best-case volume fraction resolution appears roughly consistent with CSD results. (E) dRL volume fraction resolution at SNR = 30 vs. a T = 0.6 target. Note that dRL tends to lose out at the edge cases (20/80, 30/70 splits) compared to CSD, though this is improved by increasing the iteration count. (F) dRL volume fraction resolution at SNR = 50 vs. a T = 0.3 target.