TABLE V. Comparison of Our Work With Previous Works.
Year | Authors | No. of seizurein original dataset | Features | Classifier | Sensitivity(%) | FPR/h | Augmentation technique |
---|---|---|---|---|---|---|---|
CHB-MIT | |||||||
2017 | Truong et al. [8] | 64 | STFT | CNN | 81.2 | 0.16 | Windowing |
2020 | Zhang et al. [16] | 156 | Common spatialfeatures | CNN | 92 | 0.12 | Division of a data sample into 3smaller segments and generation of artificial sample from random concatenation of these segments |
2020 | This work | 64 | STFT | CESP | 96 | 0.05 | Samples generated form DCGAN |
Epilepsyecosystem | |||||||
2020 | Stojanovic et al. [17] | 692 | Non-negativematrix factorization | SVM | 69 | 0.76 | Synthetic monitory over-sampling |
2020 | Ramy et al. [32] | 1326 | Continuous wavelettransform | Semi-DilatedCNN | 89.52 | N/M | Division of data into smaller segments |
2020 | This work | 1326 | STFT | CESP | 92.87 | 0.15 | Samples generated form DCGAN |