Table 2.
CNN + RNN classification results, evaluated on test days for each participant (Testing dataset and training dataset).
| Subject id | Sensitivity % | Specificity % | Accuracy % | Number of correctly predicted 10 minutes intervals/total | |||
|---|---|---|---|---|---|---|---|
| 5-min ECG | 10-min voting | 5-min ECG | 10 min voting | 5-min ECG | 10 min voting | ||
| Classification results evaluated on the testing days | |||||||
| Subject 1 | 79.7 | 80.5 | 69.4 | 73.3 | 73.3 | 76 | 61/82 = 74.4% |
| Subject 2 | 81.8 | 81.8 | 82.2 | 88.0 | 82.2 | 87.5 | 154/178 = 86.5% |
| Subject 3 | 82.4 | 76.5 | 89.6 | 94.6 | 89.2 | 93.7 | 166/179 = 92.7% |
| Subject 4 | 100 | 100 | 81.1 | 82.0 | 84.8 | 85.6 | 128/153 = 83.6% |
| Average | 86.0 ± 9.4 | 84.7 ± 10.4 | 80.6 ± 8.3 | 84.5 ± 9.0 | 82.4 ± 6.7 | 85.7 ± 7.3 | 84.3% |
| CNN + RNN classification results evaluated on the training days. | |||||||
| Subject 1 | 74.8 | 75.2 | 84.6 | 88.6 | 80.0 | 82.3 | 78/95 = 82.1% |
| Subject 2 | 85.9 | 84.1 | 92.8 | 96.3 | 91.4 | 94.1 | 170/180 = 94.4% |
| Subject 3 | 100 | 100 | 84.6 | 89.2 | 87.2 | 91.1 | 240/267 = 89.9% |
| Subject 4 | 92.6 | 90.2 | 91.9 | 94.6 | 92.2 | 92.6 | 145/158 = 91.8% |
| Average | 88.3 ± 10.6 | 87.4 ± 10.4 | 88.5 ± 4.4 | 92.2 ± 3.8 | 87.7 ± 5.5 | 90.0 ± 5.2 | 89.6% |
The 5-min column presents the results for the 5-minutes ECG segments used during training, whereas the 10-minute voting column shows the results when taking the majority class corresponding to all heartbeats in a 10-minute window of time.