Estimation of cardiac output and systemic vascular resistance using a multivariate regression model with features selected from the finger photoplethysmogram and routine cardiovascular measurements

Table 2 CO and SVR estimation performance and selected features

Estimated variable	r	Bias	s.d.	% error	Feature pool
CO	0.45	0	1.49	51%	All features
(L min^-1)	0.54	-0.01	1.38	47%	Only PPG features (i.e., all except 4-6, 11-13, 18-20,25-27)
	0.42	0	1.49	51%	Only PW
	0.39	-0.01	1.51	52%	Exclude PPGV (i.e., all except 1-3, 8-10, 15-17, 22-24)
	0.30	-0.01	1.62	55%	Only linear features (exclude 8-28)
SVR	0.50	1.45	236	50%	All features
(dyn.s.cm^-5)	0.55	6.50	211	45%	Only PPG features (i.e., all except 4-6, 11-13, 18-20, 25-27)
	0.14	-6.30	256	54%	Only PW
	0.40	3.17	233	49%	Exclude PPGV (i.e., all except 1-3, 8-10, 15-17, 22-24)
	0.52	-7.28	217	46%	Only linear features (exclude 8-28)

The performance figures achieved by the feature selection algorithm to estimate CO and SVR using different starting feature subsets drawn from the entire feature pool. The correlation, Bland-Altman bias and s.d., as well as the percentage error achieved is shown.

ISSN: 1475-925X