Comparative Analysis of Classifiers for Developing an Adaptive Computer-Assisted EEG Analysis System for Diagnosing Epilepsy

Table 8

Classification rate improvement caused by retraining. First column shows the channel label, the second column shows the initial training accuracy, and third one shows the marked correction by a neurologist and the last one shows the final accuracy.

Channel	Accuracy after initial training (%)	Number of epochs marked by the user	Accuracy after retraining (%)
Fp1	84.1736	57	84.6666
Fp2	85.9321	57	86.2142
F3	89.3863	57	90.1669
F4	82.5886	57	84.0065
C3	84.6526	57	85.6825
C4	84.0094	57	87.0688
P3	82.8481	57	84.6240
P4	83.0264	57	85.0625
O1	84.6445	57	84.1711
O2	83.3906	57	84.7017
F7	85.0919	57	88.3044
F8	84.8240	57	86.7992
T3	87.8007	57	88.8983
T4	92.0082	57	92.0256
T5	83.5051	57	86.4678
T6	85.5538	57	87.3744
FZ	83.2279	57	83.9403
CZ	80.0350	57	82.1546
PZ	90.5629	57	91.4475
E	87.3452	57	86.5264
PG1	88.4426	57	88.3773
PG2	83.0150	57	83.3391
A1	84.1149	57	85.5487
A2	84.0726	57	85.0169
T1	81.1019	57	83.0590
T2	83.5085	57	86.0844
X1	85.4925	57	87.7204
X2	85.2431	57	86.6259
X3	78.6937	57	80.5123
X4	81.2826	57	81.3851
X5	73.0822	57	77.2667
X6	83.9101	57	84.7386
X7	76.2795	57	79.2632