Table 1.
Row | Implementation | Avg_L a | Gauss b | 7 Peaks c | 1 Peak d | AvgError e |
---|---|---|---|---|---|---|
1 | Trace_L_G_7 | ✔ | ✔ | ✔ | ✖ | 1.300 |
2 | Trace_L_7 | ✔ | ✖ | ✔ | ✖ | 1.345 |
3 | Trace_G_7 | ✖ | ✔ | ✔ | ✖ | 1.523 |
4 | Trace_L_1 | ✔ | ✖ | ✖ | ✔ | 2.157 |
5 | Trace_L_G_1 | ✔ | ✔ | ✖ | ✔ | 2.420 |
Improvement in the seven-peak convolution f | 46.28% | |||||
Improvement in the longitudinal average g | 14.62% |
a Longitudinal average along the direction of the bundle; b isotropic Gaussian filter applied to the original density map; 2D convolution using c seven Gaussian peaks and d one Gaussian peak; e average cross-distance (in voxels) between the manually and computationally detected 330 filaments. A check mark indicates that a technique in the column was applied. For example, 1.300 voxels was the average cross-distance when the longitudinal average, isotropic Gaussian filter and seven-peak convolution were applied. f Improvement = difference in error between row 5 and row 1/error in row 5; g improvement = difference in error between row 3 and row 1/error in row 3.