Table 8.

Mutual Information of patient demographic data, PPG optical features and features derived using ECG, for the Aurora-BP dataset⁶⁰. While all features combined have an Info-Fraction of just 9.8% for the SBP prediction task, they encode much more information for the HR prediction (87.7%) and RWAT prediction (64.6%) tasks.

Feature	Mutual Information
	SBP (bits)	HR (bits)	RWAT (bits)
Age	0.026	0.015	0.000
Weight	0.024	0.024	0.000
Height	0.007	0.000	0.000
HR	0.130	2.000	0.658
std HR	0.009	1.230	0.509
rPAT	0.200	0.220	0.116
HRV	0.170	0.295	0.131
Inv. PAT	0.070	0.132	0.065
Quality	0.100	0.080	0.000
$\frac{\mathit{dp}}{\mathit{dt}}$	0.016	0.476	0.584
std $\frac{\mathit{dp}}{\mathit{dt}}$	0.009	0.232	0.252
$\mathrm{\Delta }$ rPAT	0.160	0.165	0.074
$\mathrm{\Delta }$ Inv. PAT	0.060	0.072	0.063
$\mathrm{\Delta }$ $\frac{\mathit{dp}}{\mathit{dt}}$	0.014	0.131	0.139
$\mathrm{\Delta }$ HRV	0.170	0.130	0.016
$\mathrm{\Delta }$ HR	0.025	0.242	0.234
$\mathrm{\Delta }$ Quality	0.070	0.060	0.001
Combined	0.364	3.240	1.650
Entropy	3.680	3.650	2.540
Info-Fraction (Combined)	9.89%	88.8%	65.0%