• Gaurav Kapur, Lingguang Chen, Yong Xu, Katherine Cashen, Jeff Clark, Xiaoce Feng, and Sean F Wu.
• Department of Pediatrics, Children's Hospital of Michigan/Wayne State University, Detroit, MI.
• Pediatr Crit Care Me. 2019 Sep 1; 20 (9): 809-816.

ObjectivesTo develop a novel device to predict systolic and diastolic blood pressure based on measured heart sound signals and evaluate its accuracy in comparison to intra-arterial blood pressure readings.Study DesignProspective, observational pilot study.SettingPICU.PatientsCritically ill children (0-18 yr) undergoing continuous blood pressure monitoring via radial artery intra-arterial catheters were enrolled in the study after informed consent. The study included medical, cardiac, and surgical PICU patients.InterventionsAlong with intra-arterial blood pressure, patient's heart sounds were recorded simultaneously by a highly sensitive sensor taped to the chest. Additional hardware included a data acquisition unit and laptop computer. Subsequently, advanced signal processing technologies were used to minimize random interfering signals and extract and separate S1 and S2 signals. A computerized model was then developed using artificial neural network systems to estimate blood pressure from the extracted heart sound analysis.Measurements And Main OutcomesWe found a statistically significant correlation for systolic (r = 0.964; R = 0.928) and diastolic (r = 0.935; R = 0.868) blood pressure readings (n = 491) estimated by the novel heart-sound signal-based method and those recorded by intra-arterial catheters. The mean difference of the individually paired determinations of the blood pressure between the heart-sound-based method and intra-arterial catheters was 0.6 ± 7 mm Hg for systolic blood pressure and -0.06 ± 5 mm Hg for diastolic blood pressure, which was within the recommended range of 5 ± 8 mm Hg for any new blood pressure devices.ConclusionsOur findings provide proof of concept that the heart-sound signal-based method can provide accurate, noninvasive blood pressure monitoring.

20 keywords...

hide…