Table 4 Summary of intelligent ML-based SAA using augmented signals
Refs., Year | Database | Signal for FE, Segment length | Method description | ICF | FS | FV | Limitation | Key Findings |
|---|---|---|---|---|---|---|---|---|
[44], 2017 | CUDB, VFDB | ECG segment, NSH signal using MVMD, 8Â s | Reconstruction of NSH signal using MVMD. | 54 | GA and SFFS using 3Â ML classifiers and fivefolds CV. | Fivefolds CV | Limited database. | FFC of 20 features with SVM. |
Use of both stand-alone ECG and NSH signals for FE. | Time-comsuming for FS. | Effective twolayered FS. | ||||||
The two-layered FS with 3Â ML classifiers to select 3 CFCs. | Â | NSH signal using MVMD. | ||||||
Validation of CFCs and a combination of all ICFs using 5-folds CV. | Â | Expansion of new ICFs. | ||||||
[45], 2018 | CUDB | ECG segment, (ECG segment)\(^2\) using Stationary wavelet transform, 4.1Â s | Calculation of sample entropy for 10 selected bands using Stationary wavelet transform. | 24 | Bandwidth ranking with score | NA | Limited database. | Short segment. |
Bands ranked by NT-score which is a combination of relief, gain ratio, and fisher score. | No validation. | Expansion of ICFs from ECG segment and square of ECG segment. | ||||||
Performance of 3Â ML classifiers for VFVT/non-VFVT and VF/non-VF scenarios. | Limited number of ICFs | Â |