Iterative Run-to-Run Learning Model to Derive Continuous Brachial Pressure Estimates from Arterial and Venous Lines During Dialysis Treatment

Stewart, Jill; Stewart, Paul; Walker, Tom; Viramontes Horner, Daniela; Lucas, Bethany; White, Kelly; Taal, Maarten W.; Selby, Nicholas M.; Morris, Mel

Abstract

Objective: Non-invasive continuous blood pressure monitoring is not yet part of routine practice in renal dialysis units but could be a valuable tool in the detection and prevention of significant variations in patient blood pressure during treatment. Feasibility studies have delivered an initial validation of a method which utilises pressure sensors in the extra-corporeal dialysis circuit, without any direct contact with the person receiving treatment. Our main objective is to further develop this novel methodology from its current early development status to a continuous-time brachial artery pressure estimator. Method: During an in vivo patient feasibility study with concurrent measurement validation by Finapres Nova experimental physiological measurement device, real-time continuous dialysis line pressures, and intermittent occluding arm cuff pressure data were collected over the entire period of (typically 4-hour) dialysis treatments. There was found to be an underlying quasilinear relationship between arterial line and brachial pressure measurements which supported the development of a mathematical function to describe the relationship between arterial dialysis line pressure and brachial artery BP. However, unmodeled non-linearities, dynamics and time-varying parameters present challenges to the development of an accurate BP estimation system. In this paper, we start to address the problem of physiological parameter time variance by novel application of an iterative learning run-to-run modeling methodology originally developed for process control engineering applications to a parameterised BP model. Results: The iterative run-to-run learning methodology was applied to the real-time data measured during an observational study in 9 patients, supporting subsequent development of an adaptive real-time BP estimator. Tracking of patient BP is analysed for all the subjects in our patient study, supported only by intermittent updates from BP cuff measurements. Conclusion: The methodology and associated technology is shown to be capable of tracking patient BP non-invasively via arterial line pressure measurement during complete 4-hour treatment sessions. A robust and tractable method is demonstrated, and future refinements to the approach are defined.

Citation

Stewart, J., Stewart, P., Walker, T., Viramontes Horner, D., Lucas, B., White, K., …Morris, M. (2021). Iterative Run-to-Run Learning Model to Derive Continuous Brachial Pressure Estimates from Arterial and Venous Lines During Dialysis Treatment. Biomedical Signal Processing and Control, 65, Article 102346. https://doi.org/10.1016/j.bspc.2020.102346

• Journal Article Type: Article
• Acceptance Date: Nov 15, 2020
• Online Publication Date: Nov 28, 2020
• Publication Date: 2021-03
• Deposit Date: Nov 16, 2020
• Publicly Available Date: Nov 29, 2021
• Journal: Biomedical Signal Processing and Control
• Print ISSN: 1746-8094
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 65
• Article Number: 102346
• DOI: https://doi.org/10.1016/j.bspc.2020.102346
• Public URL: https://nottingham-repository.worktribe.com/output/5047576