Table B1.
Classifiers | % classification error | % TP (nuclei) |
---|---|---|
SVM order = 3 | 42.10 | 9.48 |
SVM order = 5 | 29.57 | 18.22 |
SVM order = 7 | 34.89 | 29.84 |
Random forest (100) | 13.87 | 43.18 |
Random forest (500) | 13.53 | 60.95 |
Naive‐Bayes | 29.31 | 66.58 |
Note: SVM performances are very poor, in particular the generalisation for minor classes. RF has a relatively low misclassification error but the TP rate is not sufficient to accurately identify thalamic nuclei. NB results in bad classifications, although TP rate for nuclei class is the highest among the rejected classifiers.