Table 4A.
One-way ANOVA test on the classification performance across different threshold values for nine tractography algorithms in three diagnostic tasks.
| Degrees of freedom | Global threshold |
Individual binary threshold |
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AD vs. NC |
AD vs. MCI |
MCI vs. NC |
AD vs. NC |
AD vs. MCI |
MCI vs. NC |
|||||||||
| F | Sig. | F | Sig. | F | Sig. | F | Sig. | F | Sig. | F | Sig. | |||
| Tensor-FACT | Between groups | 7 | 0.410 | 0.895 | 0.578 | 0.773 | 1.195 | 0.309 | 0.507 | 0.828 | 1.391 | 0.213 | 0.474 | 0.852 |
| Within groups | 152 | |||||||||||||
| Tensor-RK2 | Between groups | 7 | 0.096 | 0.998 | 0.245 | 0.973 | 0.924 | 0.490 | 1.176 | 0.320 | 2.336 | 0.027 | 1.020 | 0.419 |
| Within groups | 152 | |||||||||||||
| Tensor-SL | Between groups | 7 | 0.100 | 0.998 | 2.055 | 0.052 | 0.188 | 0.988 | 1.133 | 0.346 | 0.566 | 0.783 | 0.970 | 0.456 |
| Within groups | 152 | |||||||||||||
| Tensor-TL | Between groups | 7 | 0.378 | 0.914 | 0.072 | 0.999 | 0.525 | 0.815 | 0.804 | 0.585 | 2.285 | 0.031 | 2.474 | 0.020 |
| Within groups | 152 | |||||||||||||
| ODF-FACT | Between groups | 7 | 0.752 | 0.628 | 0.412 | 0.894 | 0.645 | 0.718 | 0.455 | 0.865 | 0.608 | 0.749 | 0.250 | 0.972 |
| Within groups | 152 | |||||||||||||
| ODF-RK2 | Between groups | 7 | 2.393 | 0.024 | 1.030 | 0.412 | 0.302 | 0.952 | 1.445 | 0.191 | 1.030 | 0.413 | 1.062 | 0.391 |
| Within groups | 152 | |||||||||||||
| Probtrackx | Between groups | 7 | 1.410 | 0.205 | 0.481 | 0.847 | 0.279 | 0.962 | 0.423 | 0.887 | 0.727 | 0.649 | 0.591 | 0.763 |
| Within groups | 152 | |||||||||||||
| PICo | Between groups | 7 | 0.138 | 0.995 | 0.153 | 0.993 | 0.579 | 0.772 | 0.572 | 0.778 | 0.916 | 0.496 | 4.114 | 0.000 |
| Within groups | 152 | |||||||||||||
| Hough | Between groups | 7 | 0.289 | 0.957 | 0.887 | 0.518 | 0.506 | 0.829 | 0.183 | 0.988 | 3.049 | 0.005 | 2.538 | 0.017 |
| Within groups | 152 | |||||||||||||
The F column is the computed F statistic and the “Sig.” column shows the p-value. Only cells with Sig. value < 0.05 are treated as nominally significant. (If “Sig. = 0.000” is written, this value is less than 0.001). Since we have eight threshold values (0.05-0.4 in intervals of 0.05), the number of degrees of freedom for the Between Groups comparison is 8-1 = 7. Moreover, since we have 20 splits for each threshold value, the number of degrees of freedom for the Within Groups comparison is 20x8-8 = 152. Thus, our critical F-value for a number of degrees of freedom = (7,152) at α = 0.05 is 2.0703. Our null hypothesis, H0, was that there is no significant difference among different threshold values, so we would only reject H0 when our computed F-value > 2.0703. There were three tests with large F-values (>2.0703) but they did not pass the Bonferroni correction in post hoc comparisons. These cells included the ODF-RK2 method in the task AD vs. NC using the “Global Threshold,” tensor-TL in task AD vs. MCI using the Individual Binary Threshold, and the Hough method in task MCI vs. NC using Individual Binary Threshold. We treated these three cases as not significant. The rest of the cases with large F-values (>2.0703) are marked as red in the table and the corresponding post hoc multiple group comparisons in these cases are presented in (B).