Table 5.
Various recent methods for identifying PVC.
| Classifier Type/Approach | Features | Acc. (%) | Spec. (%) | Pre. (%) | Rec. (%) |
F1-Score (%) |
|---|---|---|---|---|---|---|
| 2D CNN (Proposed-T1) 2D CNN (Proposed-T2) 2D CNN (Proposed-T3) 2D CNN (Proposed-T4) |
Transformation of time-series ECG data/signal into the respective 2D beat images | 99.74 99.89 99.93 99.77 |
99.58 99.85 99.94 99.73 |
97.64 99.33 99.49 99.21 |
99.38 99.80 99.93 99.65 |
99.18 99.75 99.84 99.64 |
| 2D CNN [70] | Time frequency images | 97.89 | 97.17 | 98.58 | --- | --- |
| 2D CNN [71] | Wavelet power spectrums | 97.96 | 99.11 | 82.60 | --- | --- |
| 2D CNN [72] | Wavelet fusion method, Tucker-decomposition | 90.84 | 99.86 | 78.60 | --- | --- |
| DNN [73] | R-peak amplitude, S-peak amplitude, R-R interval time, Q-peak amplitude, ventricular activation time, QRS duration time | 99.41 | --- | 96.08 | --- | --- |
| DNN [6] | Seven statistical and three morphological features | 98.60 | --- | 98.70 | --- | --- |
| Adaptive Thresholding [74] | Energy wavelet coefficients | --- | 99.94 | 99.18 | --- | --- |
| Artificial Immune Systems (AIS) [7] | Geometrical features | 98.04 | 98.65 | 91.08 | --- | --- |
| SVM [8] | Extraction of six features with several methodologies | 99.78 | 99.37 | 99.91 | --- | --- |