Table 6.

The summary of BP estimation methods based on deep learning.

Authors	Dataset	Signals	AI algorithm	Result (mmHg)
				SBP	DBP
Tazarv et al. [72]	MIMIC II 20 subjects	PPG	CNN-LSTM	MAE: 3.70 STD: 3.07	MAE: 2.02 STD: 1.76
Chuang et al. [78]	MIMIC 45 subjects	PPG, ECG	CNN-LSTM+self-attention	MAE: 2.94 STD: 4.65	MAE: 2.02 STD: 3.81
Treebupachatsakul et al. [83]	UCI 812 samples	PPG, ECG	CAN	RMSE: 7.1455	RMSE: 6.0862
Mou et al. [73]	MIMIC 3 subjects	PPG	CNN-LSTM	MAE: 4.42 for ABP
Paviglianiti et al. [81]	MIMIC 40 subjects	PPG, ECG	ResNet, LSTM, WaveNet, ResNet+LSTM	MAE: 4.118	MAE: 2.228
Slapničar et al. [21]	MIMIC III 510 subjects	PPG, derivatives	Spectrotemporal ResNet	MAE: 9.43	MAE: 6.88
Brophy et al. [85]	UCI_BP Queensland 6 subjects	PPG	GAN	MAE: 2.95 STD: 19.33 for MAP
Aguirre et al. [18]	MIMIC 1131 subjects	PPG	RNN encoder-decoder + attention	MAE: 6.57 STD: 0.20	MAE: 14.39 STD: 0.42
Wang et al. [82]	UCI_BP 348 records	Image transformed from PPG	Pretrained AlexNet, Inception-V3, VGG-19	MAE: 6.17	MAE: 3.66
Esmaelpoor et al. [74]	MIMICII 200 subjects	PPG	CNN-LSTM	MAE: 3.97 STD: 5.55	MAE: 2.10 STD: 2.84
Baker et al. [75]	MIMIC III 200000 segments	PPG, ECG	CNN-LSTM	MAE: 4.41 STD: 6.11	MAE: 2.91 STD: 4.23
Qiu et al. [77]	MIMIC 1216 subjects	PPG, ECG	ResNet + SE	MAE: 3.70	MAE: 2.81
Leitner et al. [31]	MIMIC 100 subjects	PPG	CNN-GRU	MAE: 3.52	MAE: 2.20
Schrumpf et al. [80]	MIMIC 3750 + 625 subjects	PPG	AlexNet, ResNet, LSTM, model of Slapničar et al.	MAE: 16.4	MAE: 8.5
Yen et al. [87]	UCI 1551 subjects	PPG	CNN-LSTM	MAE: 2.942 STD: 5.076	MAE: 1.747 STD: 3.042
Tanveer et al. [76]	MIMIC I 39 subjects	PPG, ECG	ANN-LSTM	MAE: 1.10	MAE: 0.58
Panwar et al. [86]	MIMIC II 1557 subjects	PPG	CNN-LSTM	MAE: 2.30 STD: 0.196	MAE: 3.97 STD: 0.064
Sadrawi et al. [84]	Self-collected 18 subjects	PPG	GA + Lenet5/U-net	MAE: 2.54	MAE: 1.48