Table 10.
Performance comparison with related works.
| Author | Data Size | Calibration | Model | Input | SBP (mmHg) | DBP (mmHg) | |
|---|---|---|---|---|---|---|---|
| Inputs | Signal | Error | Error | ||||
| Chan et al. [14] | Unspecified | Cal-based | Linear regression |
Feature (PTT) |
ECG PPG |
ME: 7.49 STD: 8.82 |
ME: 4.08 STD: 5.62 |
| Kachuee et al. [15] |
1000 subjects 10 min (MIMIC 3) |
Cal-based | AdaBoost | Features | ECG PPG |
MAE: 8.21 STD: 5.45 |
MAE: 4.31 STD: 3.52 |
| Cal-free | MAE: 11.17 STD: 10.09 |
MAE: 5.35 STD: 6.14 |
|||||
| Kurylyak et al. [17] |
15,000 heartbeats |
Cal-based | Deep learning (ANN) |
Features | PPG | ME: 3.80 STD: 3.46 |
ME: 2.21 STD: 2.09 |
| Lee et al. [13] | 30 subjects | Cal-based | Deep learning (ANN) |
Feature (IPD) |
BCG | ME: 0.01 STD: 6.75 |
ME: 0.05 STD: 5.83 |
| Slapnivcar et al. [19] |
510 subjects 700 h (MIMIC 3) |
Cal-based | Deep learning (ResNet) |
Raw | PPG | MAE: 9.43 | MAE: 6.88 |
| Cal-free | MAE: 15.41 | MAE: 12.38 | |||||
| Su et al. [16] | 84 subjects 10 min |
Cal-based | Deep learning (RNN) |
Features | ECG PPG |
RMSE: 3.73 | RMSE: 2.43 |
| Tanveer et al. [20] |
39 subjects (MIMIC 1) |
Cal-based | Deep learning (ANN+ LSTM) |
Raw | ECG PPG |
RMSE: 1.27 MAE: 0.93 |
RMSE: 0.73 MAE: 0.52 |
| Wang et al. [18] | 58,795 intervals of PPG (MIMIC 1) |
Cal-based | Deep learning (ANN) |
Features | PPG | MAE: 4.02 STD: 2.79 |
MAE: 2.27 STD: 1.82 |
| This study | 15 subjects 30 min |
Cal-based | Deep learning (CNN+ Bi-GRU) |
Raw | BCG | ME: −0.82 STD: 7.50 |
ME: −0.97 STD: 5.36 |
| ECG PPG |
MAE: 4.46 STD: 4.06 |
MAE: 3.70 STD: 3.37 |
|||||
| Deep learning (CNN+ Bi-GRU+ Attention) |
ECG PPG BCG |
MAE: 4.06 STD: 4.04 |
MAE: 3.33 STD: 3.42 |
||||