Research Article
ECG Classification for Detecting ECG Arrhythmia Empowered with Deep Learning Approaches
Table 7
Simulation result of proposed CAA-TL model.
| | AlexNet | SqueezeNet | ResNet50 | | Image dimensions | 227 × 227 | 227 × 227 | 227 × 227 | | Layers | 25 | 68 | 177 |
| | For F | Training | Validation | Training | Validation | Training | Validation | | Accuracy | 97.38% | 76.67% | 77.19% | 79.20% | 77.19% | 79.25% | | Miss classification rate | 2.62% | 3.23% | 22.81% | 20.80% | 22.81% | 20.75% | | Sensitivity | 92.40% | 91.30% | 44.62% | 43.53% | 44.62% | 43.60% | | Specificity | 98.32% | 97.34% | 100% | 58.21% | 100% | 100% | | Precision | 91.17% | 91.08% | 100% | 99.83% | 100% | 100% | | FPR | 0.02% | 0.03% | 0% | 0.99% | 0% | 0% | | FNR | 0.08% | 0.09% | 0.55% | 0.56% | 0.55% | 0% |
| | For S | Training | Validation | Training | Validation | Training | Validation | | Accuracy | 97.70% | 97.50% | 100% | 71.55% | 100% | 71.55% | | Miss classification rate | 2.30% | 2.5% | 0% | 28.45% | 0% | 30.66% | | Sensitivity | 95.66% | 94.52% | 100% | 42.92% | 100% | 42.92% | | Specificity | 98.24% | 98.42% | 100% | 100% | 100% | 100% | | Precision | 93.5% | 94.88% | 100% | 100% | 100% | 100% | | FPR | 0.02% | 0.02% | 0% | 0% | 0% | 0% | | FNR | 0.04% | 0.06% | 0% | 0.57% | 0% | 0% |
| | For Q | Training | Validation | Training | Validation | Training | Validation | | Accuracy | 98.85% | 98.96% | 100% | 99.97% | 100% | 100% | | Miss classification rate | 1.15% | 1.04% | 0% | 0.027% | 0% | 0% | | Sensitivity | 96.73% | 97.61% | 100% | 100% | 100% | 100% | | Specificity | 99.17% | 99.28% | 100% | 99.97% | 100% | 100% | | Precision | 94.6% | 95.8% | 100% | 99.8% | 100% | 100% | | FPR | 0.008% | 0.007% | 0% | 0.0003% | 0% | 0% | | FNR | 0.03% | 0.02% | 0% | 0% | 0% | 0% |
| | For N | Training | Validation | Training | Validation | Training | Validation | | Accuracy | 99.47% | 99.74% | 77.19% | 83.96% | 77.19% | 79.25% | | Miss classification rate | 0.53% | 0.29% | 22.81% | 16.04% | 22.81% | 20.75% | | Sensitivity | 98.34% | 99.17% | 0% | 0% | 0% | 0% | | Specificity | 99.76% | 99.87% | 100% | 83.96% | 100% | 79.25% | | Precision | 99.10% | 99.58% | 0% | 0% | 0% | 0% | | FPR | 0.002% | 0.001% | 0% | 0.16% | 0% | 0.21% | | FNR | 0.017% | 0.008% | 1% | 1% | 1% | 1% |
| | For V | Training | Validation | Training | Validation | Training | Validation | | Accuracy | 99.26% | 97.48% | 100% | 71.52% | 100% | 71.55% | | Miss classification rate | 1.74% | 2.52% | 0% | 28.48% | 0% | 28.45% | | Sensitivity | 95.62% | 93.77% | 100% | 0% | 100% | 0% | | Specificity | 99.36% | 98.44% | 100% | 71.54% | 100% | 71.55% | | Precision | 98.40% | 94% | 100% | 0% | 100% | 0% | | FPR | 0.006% | 0.02% | 0% | 0.28% | 0% | 0.28% | | FNR | 0.04% | 0.06% | 0% | 1% | 1% | 1% |
|
|
