The Project: Ensure the algorithm for ECG analysis is enabled to be flexible in order for it to be able to be used in an array of devices.
The Solution: Use an MCU that offers ultra low power consumption in consideration of wireless and handheld devices.
By Avi Wolfson and David Niewolny
Figure 1: ECG Signal containing fudical points and measurements.
Cardiovascular disease is the leading cause of death globally. According to the World Health Organization, approximately 17.5 million people died from cardiovascular disease in 2005, or about 30% of all global deaths. Of these deaths, 7.6 million were caused by heart attacks, and 5.7 million were due to stroke. By 2015, an estimated 20 million people are expected to die from cardiovascular disease every year, primarily from heart attacks and strokes. Many of these deaths may occur with no previous symptoms of cardiovascular disease.
Electrocardiogram (ECG) monitors are vital tools used by healthcare providers to help identify cardiac conditions and monitor patient health. The ECG monitors the heart by measuring the voltages generated by the cells in the heart. The cardiac electrical vector is generated in the right atrium of the sinuatrial node (SA) and transmitted as a vector to the entire heart muscle. The measured results are graphically plotted on a y-axis against time on the x-axis, giving the signal shown in Figure 1.
An ECG monitor can come in many forms, ranging from handheld devices for remote monitoring to sophisticated patient monitoring systems used by hospitals. Though the form factors and applications are different, the basic design of any ECG device has been well established since Willem Einthoven constructed the first practical ECG in the early twentieth century.
To acquire the ECG signal, skin sensors must be applied to the patient at various pre-prescribed locations. As the heart goes through its normal cardiac cycle, the electrical activity recorded at the surface of the sensors changes. Since heart beats in a healthy patient all consist of a standard series of cardiac depolarization and repolarization events, the ECG profile of such a patient exhibits strong periodicity. A patient with an irregular heart beat, such as a heart flutter or arrhythmia, on the other hand, can produce an ECG profile with aperiodic features. Alternatively, an irregular heart beat can manifest as an ECG that is different from a healthy patient, although still periodic. Examples of this irregularity include overly slow or rapid heart rates, in which each beat still resembles every other beat but occurs with a different frequency than found in a healthy heart.
The novelty of recent ECG device designs lies not in their form but in their function. While older ECG machines would print out the electrical profile for clinical interpretation, many modern versions use advanced algorithms to monitor and diagnose patients without the need for a clinician to be present. Although clinical assessment of the ECG profile is still very important in many situations, automating the process allows a greater number of patients to be monitored for a greater amount of time. This type of automation has enabled the advent of cardiac event monitors that don’t require constant clinical oversight but instead alert clinicians if and when an arrhythmia is detected.
Freescale and Monebo's ECG solution can be used in a wide range of medical applications such as event monitors, telehealth devices, and patient monitoring systems.
At the heart of these issues is the efficiency of the algorithm used to analyze the ECG signals and detect algorithms. Traditionally, arrhythmias have been diagnosed by highly trained clinicians. Converting the clinical eye into an electronic device is an arduous, imperfect task. However, developing novel algorithms that enable higher degrees of diagnosis accuracy and precision offers substantial payoffs. The care that patients receive depends on the device’s ability to accurately diagnosis arrhythmias each time they occur. The better an algorithm performs these tasks, the more appealing it is in a clinical setting. Unfortunately, increasing precision and accuracy generally requires greater amounts of code and greater programming complexity. This, in turn, requires the device running the code to have more on-board memory, which leads to a rise in manufacturing costs.
An ideal algorithm uses mathematical rigor to reduce code size, memory requirements, and ultimately, cost, while still beating current industry standards for accurate and precise diagnoses. Recently, Monebo software produced an example of just such an algorithm. In contrast to most commercially available algorithms, Monebo’s Kinetic Intervals ECG algorithm does not require a warm-up period of sampling a patient’s ECG profile, which would then normally be discarded. Warm-up sampling is a common practice because it ensures that the device will be reading a “true” ECG profile rather than one distorted by transients associated with starting up the device. Warm-up sampling minimizes the occurrence of false positive diagnoses. However, this approach requires extra code that is added to determine when actual testing is to begin and how and where to store the data to be discarded. The Monebo solution, on the other hand, takes advantage of advanced signal averaging techniques and sampling of key “fiducial” points to negate the need for this warm-up time. As a result, the Monebo algorithm has leaner code requirements.
Having efficient ECG detection code is a boon to device designers. It allows them to focus on specific implementations, secure in the knowledge that their arrhythmia detection algorithm will work without consuming excess memory. As was previously stated, all cardiac monitoring devices are not created equal. A simple, wireless at-home ECG monitor with few leads will require long battery life and efficient production costs, whereas a more complex, hospital-grade device may need a greater degree of processing power and a real-time operating system (RTOS) to run additional applications. Many manufacturers produce devices that span this spectrum. It becomes important, then, that the ECG algorithm can be easily ported to a range of processing cores that match the needs of the device in terms of power and capability.
Monebo has addressed this need for hardware platform flexibility by partnering with Freescale Semiconductor. By taking advantage of Freescale’s portfolio of more than 500 microcontrollers (MCUs), processors, and digital signal controllers (DSCs), Monebo has positioned their Kinetic Intervals algorithm to work across a broad spectrum of applications. In particular, Freescale’s Flexis QE family of pin- and software-compatible 8-bit and 32-bit MCUs was designed to offer ultra low power consumption for enhanced battery life. This low-power operation is a key consideration for next-generation wireless and handheld healthcare devices.
ECG technology has come a long way since Einthoven’s string galvanometer. Recent innovations in automation have allowed diagnosis of arrhythmias to benefit more people than ever before. Although great strides have been made, developers of the next generation of cardiac monitors must keep an eye on diagnosis accuracy, code complexity, economic efficiency, and the device’s power consumption. Monebo’s and Freescale’s combined software and hardware platform addresses these challenges head on and provides a collaborative solution that will serve the technology needs of ECG device developers for years to come.
For additional information on the technologies and products discussed in this article, see MDT online at www.mdtmag.com  and the following websites: