Accurate Monitoring of Cardiovascular Activity

Accurate Monitoring of Cardiovascular Activity
ECG measurement containing a motion-induced artifact

Abdelhak M Zoubir and Michael Muma of Germany’s Technische Universität Darmstadt, discuss robust signal processing and accurate monitoring of cardiovascular activity in the service of human wellbeing.

Signal processing encompasses the fundamental theory, applications, algorithms, and implementations of processing information contained in signals. Examples of signals are numerous and include medical, biological, audio, video, image, various types of sensor outputs, communication, geophysical, radar, chemical, molecular, and genomic.

Our body constantly generates, senses, analyses and manipulates signals. The cardiovascular system is powered by the heart, which pumps blood through the arteries, veins, and capillaries traversing the entire human body. The monitoring and analysing of cardiovascular signals is of great importance for the assessment of human wellbeing. In recent years, new sensing technologies and algorithms have allowed the acquisition and analysis of cardiovascular signals. Advanced clinical devices, as well as consumer devices, such as smartphones, have a great potential both for improving public awareness of health metrics and for the early diagnosis of cardiac symptoms.

According to the WHO, cardiovascular diseases (CVDs) are the number one cause of death globally. In 2015, an estimated 17.7 million people died from CVDs, representing 31% of all global deaths. Atrial fibrillation (AF) is the most common type of arrhythmia, and it is one of the major causes of stroke, heart failure, sudden death, and cardiovascular morbidity.1

Robust signal processing

Robust signal processing, broadly, involves making inference based on observations of signals that have been distorted or corrupted in some unknown manner.2 Classical statistical signal processing relies strongly on a statistical model, such as the normal (Gaussian) distribution, to describe signals and/or noise.

The Gaussian distribution is in many situations a reasonable model to describe, for example, sensor noise. It also allows the derivation of optimal methods in closed form. Robust statistical methods account for the fact that the postulated models for the data are fulfilled only approximately and not exactly. In contrast to classical procedures, robust methods are not significantly affected by small changes in the data, such as outliers or small model departures.

This comes at a price of a small performance loss when the assumptions hold exactly, i.e. robust methods are near-optimal, not optimal. While optimality is clearly desirable, robustness is the appropriate choice for cardiovascular monitoring. This article gives an insight into various problems we encountered and solutions that we developed when monitoring and analysing cardiovascular signals.

Removing Motion Artifacts in the ECG

The ECG is routinely used to monitor the electrical and muscular functions of the heart. Analysing the heart rate and its rhythm enables diagnosis and can help preventing, or treating CVDs. In a clinical setting, a patient has to hold still to obtain an accurate ECG. However, there exist situations, where this is impossible, e.g., long term monitoring, or ECG recordings of children.

In such cases, the ECG contains motion artifacts, as illustrated in Fig. 1. Robust signal processing algorithms are beneficial in these situations. For example, we designed a robust signal processing algorithm in the so-called wavelet domain, which facilitates the separation of the artifacts and the artifact-free ECG signal.3

Photoplethysmography (PPG)

Since the ECG acquires the heart’s electrical signals through electrodes on the skin, measurement devices can be inconvenient to wear in many real-life situations. This motivates optical sensing techniques, such as PPG, which are easily integrated into wearables or smartphones.

For smartphone-based recordings, one places a finger on the phone’s camera. The smartphone light illuminates the blood vessels and acquires a diffuse reddish video. As the blood flow pulsates, the red colour changes slightly and these changes can be attributed to heart’s activity. Specific irregularities that indicate AF also leave their traces.

We derived a PPG-based AF detection algorithm for smartphones that has a low computational cost and low memory requirements.1 For a small, clinically collected data set (326 measurements), we achieved perfect AF detection.

A further application of PPG is the accurate and reliable estimation of the heart rate. Especially during physical exercise with a wearable device, such as a watch with a PPG sensor, algorithms must deal with noisy signals that contain strong motion artifacts. We developed algorithms, which reduce motion artifacts in real-time by incorporating information from synchronously operated acceleration sensors.

One algorithm was developed by the Technische Universität Darmstadt student team who achieved first place in the international IEEE Signal Processing Cup 2015, a prestigious competition in which more than 50 teams competed. Currently, research is progressing to extract further physiological parameters, e.g., blood pressure and arterial stiffness.

The dynamics of the eye’s wavefront aberrations

The human eye is a complex dynamic system consisting of several optical elements. It is well established that the eye’s wavefront aberrations fluctuate in time. The role of cardiopulmonary signals, i.e., pulse and respiration, in these fluctuations was investigated4 and we proposed a set of tools, based on joint time-frequency analysis to acquire a detailed picture.

Intracranial pressure (ICP) signals

A common risk for patients with traumatic brain injuries is that the primary brain damage can lead to a secondary pathophysiological damage, which is usually associated with a significantly high or low ICP level. Thus, the continuous monitoring of signals, such as ICP, mean arterial blood pressure (MAP) or brain tissue oxygen level (PtiO2), has become a golden standard in neuro-intensive care units. The sensors, however are sensitive to patient movements and bed angles.

In co-operation with the Lab of Computational Physiology, Massachusetts Institute of Technology, USA, we proposed an online approach based on signal decomposition and robust statistics that enables accurate forecasting of ICP signals. We also proposed a new and robust signal processing method to assess the nonlinear interactions between ICP, MAP, and PtiO2.5

The response synchrony of physiological parameters in emotion

An emotional experience is associated with a synchrony of the physiological response parameters. We presented a new approach for the quantification of synchrony of multivariate non-stationary psychophysiological signals during emotion eliciting stimuli, which may allow the monitoring of mental wellbeing.6

This is a brief summary of the development and applications of advanced robust algorithms for the monitoring of cardiovascular activity. More details can be found under the web address below.

References

  1. Schäck, T, Harb, YS, Muma, M, & Zoubir, AM, Computationally efficient algorithm for photoplethysmography-based atrial fibrillation detection using smartphones. Proceedings of the 39th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2017 (pp. 104-108).
  2.  Zoubir, AM, Koivunen, V, Chakhchoukh, Y, & Muma, M, Robust estimation in signal processing: A tutorial-style treatment of fundamental concepts. IEEE Signal Processing Magazine, 29 (4), pp. 61-80, 2012.
  3. Strasser, F, Muma, M, & Zoubir, AM, Motion artifact removal in ECG signals using multi-resolution thresholding. Proceedings of the 20th European Signal Processing Conference (EUSIPCO), 2012 (pp. 899-903).
  4. Muma, M, Iskander, DR, & Collins, MJ, The role of cardiopulmonary signals in the dynamics of the eye’s wavefront aberrations. IEEE Transactions On Biomedical Engineering, 57(2), pp. 373-383, 2010.
  5. Schäck T, Muma, M, Feng, M, Guan, C, & Zoubir, AM, Robust nonlinear causality analysis of non-stationary multivariate physiological time series. IEEE Transactions on Biomedical Engineering (Early Access), 2017.
  6. Kelava, A, Muma, M, Deja, M, Dagdagan, JY, & Zoubir, AM, A new approach for the quantification of synchrony of multivariate non-stationary psychophysiological variables during emotion eliciting stimuli. Frontiers in Psychology, 5 (1507), pp. 1-18, 2015.

 

The Signal Processing Group

Prof Dr Ing Abdelhak M Zoubir

Dr Ing Michael Muma

zoubir@spg.tu-darmstadt.de

muma@spg.tu-darmstadt.de

www.spg.tu-darmstadt.de

This is a commercial article that will appear in SciTech Europa Quarterly issue 27, which will be published in June, 2018.

Laboratory Supplies Directory - Now Live

LEAVE A REPLY

Please enter your comment!
Please enter your name here