Table 1 The table summarizes the limitations of previous research and indicates whether the study exhibits specific pitfalls. The pitfalls are categorized into four categories: a) Data-split: Domain Overlap (denoted as D.O), Data Overlap (denoted as C.O), or Small test set (denoted as S.T). b) Over-constraining: SBP values and standard deviation (if provided) c) Unrealistic Pre-Processing: % of remaining dataset after pre-processing (if provided) d) Calibration: Correctly employed and justified for longer periods. The columns denote the presence or absence of each limitation, with Y (yes) indicating that the study has the limitation, N (no) indicating that it does not have that limitation, U (unknown) indicating that there is not enough information available, and “-” indicating that the research is not applicable to the pitfall. For additional information, please see Section Review of the results and limitations of prior work.
From: Examining the challenges of blood pressure estimation via photoplethysmogram
Method | Dataset | Results (SBP) | Data-split | Over-constraining | Unrealistic Pre-Proc. | Calibration |
|---|---|---|---|---|---|---|
BiGRU Attention30 | MIMIC-II | MAE=2.58 SD=3.35 | U | N SD=14.1 | N \(\sim\)10% | – |
AdaBoost31 | MIMIC-II | ME=0.09 MAE=8.22 SD=10.38 | N D.O | Y | Y | – |
ANN32 | MIMIC-II | MAE=3.21 RMSE=4.23 | U | Y | N \(\sim\)75% | – |
LSTM33 | MIMIC-II | MAE=3.23 STD=4.75 | U | U | Y | – |
Ensemble CNN34 | MIMIC-III | MAE=9.43 | Y S.T | Y | N \(\sim\)1.7% | – |
ANN35 | MIMIC | MAE=4.02 SD=2.79 | U S.T | U | Y | – |
Regression36 | Custom Dataset | MAE=6.90 SD=9.00 | Y S.T | Y | Y | – |
SVR37 | Queensland | ME=11.6 SD=8.20 | Y S.T | Y | Y | – |
Regression38 | Custom Dataset | MAE=3.90 SD=5.37 | Y S.T | N | Y | – |
ANN39 | MIMIC-II | ME=0.16 MAE=4.47 SD=6.85 | N C.O | U | Y | – |
SVR40 | Custom Dataset | ME=5.10 SD=4.30 | Y S.T | N SD=11.9 | Y | – |
SVR41 | Queensland | MAE=4.76 SD=7.68 | N D.O S.T | N | Y | – |
Math Models42 | Custom Dataset | MAE=7.66 | Y S.T | N SD=12.5 | Y | – |
ANN43 | MIMIC | MAE=3.80 SD=3.46 | U S.T | N | N | – |
Regression44 | MIMIC | MAE=4.90 SD=6.59 | N D.O S.T | Y | Y | N |
LSTM-CNN45 | MIMIC-II | ME=1.55 SD=5.41 | U S.T | N | N \(\sim\)15% | – |
AdaBoost46 | MIMIC-II | ME= -0.05 SD=8.90 | N D.O | Y | N \(\sim\)20% | – |
U-Net21 | MIMIC-II | ME=-1.58 SD=8.61 | N C.O | Y | Y | – |
CNN Siamese27 | MIMIC-II | MAE=5.95 SD=6.90 [Calib] | Y | N | N \(\sim\)5% | N |
U-Net29 | MIMIC-II | ME=4.30 SD=6.50 | Y | N SD=13.5 | Y | – |
1-D CNN47 | Custom Dataset | SD=11.4 | N D.O S.T | N SD=16 | Y | – |
LSTM48 | MIMIC-II | ME=4.05 SD=4.60 | N D.O S.T | N | N \(\sim\)50% | – |
