Skip to main content
. 2022 Nov 14;9(11):690. doi: 10.3390/bioengineering9110690

Figure 5.

Figure 5

In the left confusion matrix, the performance of medical evaluation (metadata’s “EEG Result”) based on a first evaluation of the EEG signal is compared with the actual diagnosis two years later. The test set is used for focal lesional epilepsy (labeled as 1) vs. non-epileptic classification (labeled as 0). The accuracy of this first medical evaluation is 0.646, the sensitivity is 0.660 and the specificity is 0.629. In the mid confusion matrix, the result of an AND operator between the medical diagnosis from the first EEG and our best algorithm (SVM and PCA) is assessed. The accuracy is 0.634, the sensitivity is 0.553 and the specificity is 0.743. In the right confusion matrix, the result of an OR operator between the EEG result and the ML prediction is shown. The accuracy achieved is 0.683, the sensitivity is 0.851 and the specificity is 0.457.