Table 6B.
Post hoc group comparisons.
| Task | (I) Tractography algorithm | (J) Tractography algorithm | Mean difference (I-J) | Sig. | 95% confidence interval |
|
|---|---|---|---|---|---|---|
| Lower bound | Upper bound | |||||
| AD vs. NC | Tensor-SL | ODF-FACT | -0.09444 | 0.006 | -0.1741 | -0.0148 |
| ODF-RK2 | -0.09028 | 0.011 | -0.1700 | -0.0106 | ||
| CMI vs. NC | Probtrackx | Tensor-FACT | 0.10109 | 0.011 | 0.0119 | 0.1903 |
| Tensor-RK2 | 0.10091 | 0.011 | 0.0117 | 0.1901 | ||
| Tensor-TL | 0.09339 | 0.030 | 0.0042 | 0.1826 | ||
| ODF-RK2 | 0.09348 | 0.030 | 0.0043 | 0.1827 | ||
The “Sig.” column show the SPSS adjusted p-value; only values 0.05 are treated as significant. Only comparisons that passed Bonferroni correction are listed here. For the task AD vs. NC, the classification performance of tensor-SL is significantly poorer than that of ODF-FACT or ODF-RK2. Interestingly, for the task MCI vs. NC, Probtrackx has statistically better performance than the four deterministic tractography algorithms (tensor-FACT, RK2, TL, and ODF-RK2).