Table 4.

SBP and DBP prediction errors (mmHg) on Polito database using the best neural network (ResNet + LSTM) trained on MIMIC dataset (built using PPG and ECG lead I)

Tested set	MAE SBP	MAE DBP	RMSE SBP	RMSE DBP
	PPG
Validation set	7.409	3.706	9.875	4.883
Leave-One-Out	15.706	7.251	17.792	8.171
Polito dataset	9.916	5.905	11.879	7.273
	PPG + ECG
Validation set	4.546	2.515	5.766	2.982
Leave-One-Out	16.128	6.743	17.875	7.902
Polito dataset	12.435	8.567	14.082	10.211