Noninvasive blood pressure monitoring from the supraorbital artery using an artificial neural network oscillometric algorithm

Objective. Our objective was to overcome the limitations of linear models of oscillometric blood pressure determination by using a nonlinear technique to model the relationship between the oscillometric envelope and systolic and diastolic blood pressures, and then to use that technique for near-continuous arterial pressure monitoring at the supraorbital artery.Methods. An adhesive pressure pad and transducer were used to collect oscillometric data from the supraorbital artery of 85 subjects. These data were then used to train an artificial neural network (ANN) to report diastolic or systolic pressure. Arterial pressure measurements defined by brachial artery auscultation were used as a reference. ANN results were compared with those obtained using a standard oscillometric algorithm that determined pressures based on fixed percentages of the maximum oscillometric amplitude.Results. The ANN produced better estimates of reference blood pressures than the standard oscillometric algorithm. Mean difference between target and actual output for the ANN was 0.50±5.73 mm Hg for systolic pressures, compared to the mean difference of the standard algorithm of 2.78±19.38 mm Hg. For diastolic pressures, the ANN had a mean difference of 0.04±4.70 mm Hg, while the mean difference of the standard algorithm was –0.34±9.75 mm Hg.Conclusions. The ANN produced a better model of the relationship between the oscillometric envelope and reference systolic and diastolic pressures than did the standard oscillometric algorithm. Noninvasive blood pressure measured from the supraorbital artery agreed with pressure measured by auscultation in the brachial artery, and may sometimes be more clinically useful than an arm cuff device.This research was supported, in part, by a grant from Baxter Healthcare Corporation (Santa Ana, CA), and Innerspace Medical (Irvine, CA). A grant of computer time from the Utah Supercomputer Institute, which is funded by the State of Utah and the IBM Corporation, is gratefully acknowledged.