The CNN classifier shows good test performance
| Model | Sensitivity % | Specificity % | Precision % | bAcc % |
|---|---|---|---|---|
| CNN CNN-rnd | 71.0 ± 5.5 4.6 ± 3.1 | 78.1 ± 3.5 95.3 ± 3.3 | 59.7 ± 2.7 30.0 ± 3.5 | 74.6 ± 1.3 49.9 ± 0.2 |
Average test performance scores are shown for four different metrics. CNN, classifier trained on expert-annotated EEG dataset; CNN-rnd, classifier trained on the same dataset where annotations were randomly generated with class probabilities equal to the ratio of examples in each class in the original dataset; bAcc, balanced accuracy. Mean ± SD values are shown. Subject-wise five-fold cross-validation was used in each case. Scores were calculated based on the probability threshold of 0.5.