Innovative continuous non-invasive cuffless blood pressure monitoring based on photoplethysmography technology
Purpose: To develop and validate a continuous non-invasive blood pressure (BP) monitoring system using photoplethysmography (PPG) technology through pulse oximetry (PO). Methods: This prospective study was conducted at a critical care department and post-anesthesia care unit of a university teaching...
| Autores: | , , , , , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Fecha de publicación: | 2013 |
| País: | España |
| Institución: | Universitat Autònoma de Barcelona |
| Repositorio: | Dipòsit Digital de Documents de la UAB |
| Idioma: | inglés |
| OAI Identifier: | oai:ddd.uab.cat:324158 |
| Acceso en línea: | https://ddd.uab.cat/record/324158 https://dx.doi.org/urn:doi:10.1007/s00134-013-2964-2 |
| Access Level: | acceso abierto |
| Palabra clave: | Non-invasive hemodynamic monitoring Blood pressure Photoplethysmography Critical care Machine learning |
| Sumario: | Purpose: To develop and validate a continuous non-invasive blood pressure (BP) monitoring system using photoplethysmography (PPG) technology through pulse oximetry (PO). Methods: This prospective study was conducted at a critical care department and post-anesthesia care unit of a university teaching hospital. Inclusion criteria were critically ill adult patients undergoing invasive BP measurement with an arterial catheter and PO monitoring. Exclusion criteria were arrhythmia, imminent death condition, and disturbances in the arterial or the PPG curve morphology. Arterial BP and finger PO waves were recorded simultaneously for 30 min. Systolic arterial pressure (SAP), mean arterial pressure (MAP), and diastolic arterial pressure (DAP) were extracted from computer-assisted arterial pulse wave analysis. Inherent traits of both waves were used to construct a regression model with a Deep Belief Network-Restricted Boltzmann Machine (DBN-RBM) from a training cohort of patients and in order to infer BP values from the PO wave. Bland-Altman analysis was performed. Results: A total of 707 patients were enrolled, of whom 135 were excluded. Of the 572 studied, 525 were assigned to the training cohort (TC) and 47 to the validation cohort (VC). After data processing, 53,708 frames were obtained from the TC and 7,715 frames from the VC. The mean prediction biases were -2.98 ± 19.35, -3.38 ± 10.35, and -3.65 ± 8.69 mmHg for SAP, MAP, and DAP respectively. Conclusions: BP can be inferred from PPG using DBN-RBM modeling techniques. The results obtained with this technology are promising, but its intrinsic variability and its wide limits of agreement do not allow clinical application at this time. |
|---|