Table 2.
The performance metrics of comparing between AI model vs clinicians.
| Precision | Recall | F1 score | Accuracy | |
|---|---|---|---|---|
| XGBoost model | 0.9269 | 0.9300 | 0.9264 | 0.9300 |
| Ensemble modela | 0.9291 | 0.9300 | 0.9269 | 0.9300 |
| Doctor without training | 0.7562 | 0.3400 | 0.3651 | 0.3400 |
| Junior neurologist | 0.7528 | 0.6600 | 0.6893 | 0.6600 |
| Senior neurologist | 0.7689 | 0.7500 | 0.7582 | 0.7500 |
Abbreviations: XGBoost, extreme gradient boosting.
Ensemble model is composed of 80% of XGBoost and 20% of TabNet.