Table 1 Summary of the threshold-based SAAs
Refs., Year | Database | Signal for FE, Segment length | Method description | Feature | Limitation | Key Findings |
|---|---|---|---|---|---|---|
[10], 2007 | MITDB, CUDB, AHA | Stand-alone ECG, 8Â s | Counting number of boxes on a grid filled by ECG and its delay signals. | Phase space reconstruction. | No validation. Low performance. | Phase space reconstruction parameter and a threshold. VF and NSH signals show irregular and regular behaviors. |
[11], 2009 | MITDB, CUDB | Stand-alone ECG, 8Â s | Assessment of the proportion of time in which the ECG signal is above a certain threshold. | Threshold crossing sample count. | No validation. Low performance. | Threshold crossing sample count parameter and a threshold. The time in which VF signal remains outside the threshold is larger than that in which NSH signal does. |
[12], 2010 | MITDB, CUDB, VFDB | Stand-alone ECG and intrinsic mode functions using EMD, 8Â s | Calculation of mean absolute value of the signal for SH/NSH rhythm classification. Calculation of differences between ECG and first 2 intrinsic mode functions. Normalized mean absolute value for VF/VT classification. | Mean absolute value and normalized mean absolute value | No validation. Requirement of 2 algorithms. Low performance. | SH signal hardly goes through the baseline and low absolute amplitude of the NSH signal for most of the time. Similarity between VF signal and its first 2 intrinsic mode functions. |
[13], 2011 | MITDB, CUDB, VFDB | Stand-alone ECG and intrinsic mode functions using EMD, 8Â s | Measurement of correlation between ECG signal and its first intrinsic mode function, residual using the angles to form a complex decision parameter. | Complex decision parameter | No validation Low performance. | Complex decision parameter and a threshold. Correlation between ECG signal and its first intrinsic mode function and residual. |
[15], 2012 | CUDB, NSRDB | Stand-alone ECG, 4Â s | Extraction of 3 features using semantic mining algorithm. Construction of thesholds based on ANOVA model. | Natural frequency. Damping coefficient. Input signal. | No validation Low performance. | Semantic mining algorithm for feature extraction. Analysis model for threshold construction. |
[14], 2012 | CUDB, VFDB | Stand-alone ECG and intrinsic mode functions using EMD, 10Â s | Intrinsic mode function with EMD. Approximate entropy of first intrinsic mode function. | NA | Consideration of first intrinsic mode function. No validation Low performance. | Approximate entropy threshold. |