Detection of ventricular premature beats based on the pressure signals of a hemodialysis machine. (January 2018)

• Record Type: Journal Article
• Authors: Holmer, Mattias; Martínez, Juan Pablo; Gil, Eduardo; Sandberg, Frida; Olde, Bo; Sörnmo, Leif
• Is Part Of: Medical engineering & physics. Volume 51(2018)
• Page Start: 49
• Page End: 55
• Publication Date: 2018-01
• Subjects: Ventricular premature beats -- Extracorporeal pressure sensors -- Cardiac pressure signal -- Fistula -- Hemodialysis -- Linear discriminant analysis; Biomedical engineering -- Periodicals; Biomedical Engineering -- Periodicals; Physics -- Periodicals; Génie biomédical -- Périodiques; Biomedical engineering; Electronic journals; Periodicals; 610.28
• DOI: 10.1016/j.medengphy.2017.11.004
• Languages: English
• ISSNs: 1350-4533

Abstract

Highlights: A novel method for VPB detection is proposed, based on the signals from the arterial and the venous pressure sensors of a hemodialysis machine. Features describing amplitude, duration, and area, combined with linear discriminant analysis, is used for classification of normal beats and VPBs. It is shown that the cardiac pressure signal is suitable for VPB detection, provided that the average cardiac pulse pressure exceeds 1 mmHg. Abstract: Monitoring of ventricular premature beats (VPBs), being abundant in hemodialysis patients, can provide information on cardiovascular instability and electrolyte imbalance. In this paper, we describe a method for VPB detection which explores the signals acquired from the arterial and the venous pressure sensors, located in the extracorporeal blood circuit of a hemodialysis machine. The pressure signals are mainly composed of a pump component and a cardiac component. The cardiac component, severely overshadowed by the pump component, is estimated from the pressure signals using an earlier described iterative method. A set of simple features is extracted, and linear discriminant analysis is performed to classify beats as either normal or ventricular premature. Performance is evaluated on signals from nine hemodialysis treatments, using leave-one-out crossvalidation. The simultaneously recorded and annotated photoplethysmographic signal serves as the reference signal, with a total of 149, 686 normal beats and 3574 VPBs. The results show that VPBs can be reliably detected, quantified by a Youden's J statistic of 0.9, for average cardiac pulse pressures exceeding 1 mmHg; for lower pressures, the J statistic drops to 0.55. It is concluded that the cardiac pressure signal is suitable for VPB detection, provided that the average cardiac pulse pressure exceeds 1 mmHg.