Table 5 Comparison of the results of the state-of-the-art
From: Epilepsy seizure prediction with few-shot learning method
Authors | Method | Database | Sensitivity (%) | FPR(/h) | SOP(min) | SPH(sec) |
|---|---|---|---|---|---|---|
2017 [32] | Phase locking value + SVM | 23Chb | 82.44 | – | 5 | 0 |
2018 [16] | Zero crossings, PSD + LSTM | 23Chb | 90 | 0.11–0.02 | 15–120 | 0 |
2018 [7] | STFT + CNN | 13Chb | 81.4 | 0.16 | 30 | 300 |
2017[33] | CSP + LDA | 24Chb | 89 | 0.39 | 120 | 0 |
2019 [34] | spectral Power + 3DCNN | 16Chb | 85.7 | 0.096 | 60 | 0 |
2018 [20] | Wavelet transform + CNN | 23Chb | 87.8 | 0.14 | 10 | 0 |
2019 [35] | Raw EEG + Bi-LSTM | 22Chb | 99.72 | 0.004 | 60 | 0 |
2020 [36] | CNN + ELM | 23Chb | 95.85 | 0.045 | – | – |
2021 [37] | STFT + RDANet | 13 Chb | 89.33 | – | – | – |
This work | Raw EEG + CNN | 15Chb | 88.49 | 0.114 | 20–25 | 300–600 |
FEW-SHOT LEARNING | 1Chb | 98.52 | 0.045 | 20–25 | 300–600 |