Table 2 Comparison of classification performance across methods.
From: VENturing into machine learning for the morphological analysis of von Economo neurons
Method | Accuracy | Precision | Recall | F1-Score | ROC-AUC | Notes |
|---|---|---|---|---|---|---|
Machine learning methods | ||||||
BART | 94.2 ± 2.1 | 0.91 | 0.96 | 0.94 | 0.97 | Bootstrap mean ± SD |
C5.0 | 93.8 ± 2.3 | 0.90 | 0.95 | 0.92 | 0.96 | Bootstrap mean ± SD |
Random forest | 95.1 ± 1.8 | 0.93 | 0.97 | 0.95 | 0.98 | Bootstrap mean ± SD |
XGBoost | 94.7 ± 2.0 | 0.92 | 0.96 | 0.94 | 0.97 | Bootstrap mean ± SD |
SVM | 92.9 ± 2.5 | 0.89 | 0.94 | 0.91 | 0.95 | Bootstrap mean ± SD |
EARTH | 93.5 ± 2.2 | 0.90 | 0.95 | 0.93 | 0.96 | Bootstrap mean ± SD |
CNN methods | ||||||
Original representation | 93.64 | 0.89 | 1.00 | 0.94 | 0.95 | Single test performance |
Diameter-enhanced | 98.18 | 0.96 | 1.00 | 0.98 | 0.98 | Single test performance |
Soma-focused | 94.55 | 0.92 | 0.98 | 0.95 | 0.96 | Single test performance |
Human expert performance | ||||||
Expert classification | 78.3 ± 12.4 | 0.72 | 0.81 | 0.76 | 0.79 | Inter-rater κ = 0.0132 |
