Stryker Capstone Design Project


Capstone Symposium Presentation_.pptx

Figure 1: The four chambers of the heart, the AV node, and the SA node11. 

Background

The human heart consists of four chambers: the left atrium, the right atrium, the left ventricle, and the right ventricle.  The atria, the heart's upper chambers, and the ventricles, the heart's lower chambers, alternately contract and relax to pump blood through the heart and to the rest of the body. The heart's electrical system is what makes this possible.  Cells in the sinoatrial (SA) node, located in the right atrium, begin the electrical signal that starts the heartbeat. The electrical signal travels down to the ventricles, causing nearby parts of the heart to contract and pump blood out of the heart. 

EKGs measure the depolarization of muscle cells. They are used to diagnose irregular electrical signals in the heart.  A monitor defibrillator is a device that analyzes heart rhythms for abnormalities and, if necessary, delivers a shock to a patient with the intention of restoring a normal heart rhythm. This device shocks all of the heart cells at once, depolarizing them and allowing them to simultaneously enter into a refractory period. This allows the pacing of the heart to normalize.  A defibrillator can deliver either synchronized electrical cardioversion or defibrillation, also known as asynchronous cardioversion.  

Figure 2: The different segments of an EKG signal, highlighting the vulnerable period12. 

An arrhythmia is an irregular heartbeat caused by a problem in the heart's electrical system.  The heart may beat too fast, too slow, or skip beats.  Synchronized cardioversion is used to treat atrial fibrillation, atrial flutter, and AV nodal tachycardia.  These conditions are not immediately life-threatening, so synchronized cardioversion shocks are usually a scheduled procedure and not an urgent emergency.  The goal is to deliver a shock between the R wave and the T wave.  Shocking at the wrong time during a synchronized cardioversion procedure can lead to cardiac arrest by inducing ventricular fibrillation.  

The vulnerable period of the EKG wave is shown in figure 2. Ventricular fibrillation and ventricular tachycardia with no pulse is an urgent emergency and requires a defibrillation shock, which can be given at any point in the cardiac cycle, unlike a synchronized cardioversion shock.  Cardioversion is typically a scheduled, in-hospital procedure for patients experiencing an abnormal atrial arrhythmia. During this procedure, a high-energy shock is delivered to the heart, timed at the peak of the QRS complex via R-wave detection. Ideally, the heart rhythm will return to normal after the shock is applied. In circumstances where an arrhythmia persists,  a second shock may be required to achieve the normal electrical activity. If an error occurs during the cardioversion procedure that results in a shock being delivered during the vulnerable, transition period in the rhythm, ventricular fibrillation would be induced in the patient. The vulnerable period of a peak containing rhythm occurs directly before the T-wave (figure 2). Ventricular fibrillation and other ventricular arrhythmias do not contain R waves, a monitor defibrillator, while in cardioversion mode, is not able to synchronize to the heartbeat to deliver a shock to the patient. In the instance that a ventricular arrhythmia is induced, the operating clinician needs to switch the device to defibrillation mode. Defibrillation is a high-energy, unsynchronized shock for patients experiencing chaotic ventricular arrhythmias. There are several problems that operating clinicians can encounter in this scenario. The clinician may not realize that the mode of the monitor defibrillator needs to be switched to properly administer an unsynchronized defibrillation shock. The clinician may not be aware of the device’s default mode after a cardioversion shock is applied. Between companies and device models, the default shock mode differs1. If not terminated with defibrillation in a timely manner, many ventricular arrhythmias can result in cardiac arrest and, in rare cases, can be fatal. The clinician is put into an unexpected, high-stress, high-risk situation in which precious time can be lost while they attempt to learn how the device works.

Clinical Need

The design need is to allow the operating clinician to administer a shock to the patient without manually changing the mode of the device. Our goal was to provide an automated way for the monitor defibrillator to classify heart arrhythmias to provide the correct treatment type.

Two Approaches

Arrhythmia Classification with Machine Learning - Classify an arrhythmia as either requiring cardioversion or defibrillation using machine learning models. 

R-Wave Detection - The presence of a QRS complex indicates an atrial arrhythmia and no QRS complex indicates a ventricular arrhythmia. 

Final Deliverable 

Figure 3: Our final algorithm

Most of the details of the final algorithm cannot be disclosed due to an NDA agreement with Stryker.  We obtained ECG data from defibrillators, extracted features from that data, and into algorithm 1 we fed in both the raw data and the features, while we just gave the extracted features to algorithms 2 and 3.  Each was multiplied by a weight, which gave each method a certain level of importance. These all were factored into our final classification of what type of shock the patient needed.  See the PowerPoint above for the metrics we measured, the final results of our algorithm, and potential future work. 

Project Summary

There are a few possible outcomes upon receiving a synchronized cardioversion shock.  Heart electrical activity may return to normal and no additional shock is necessary.  Sometimes upon receiving a cardioversion shock an atrial arrhythmia persists after the shock.  A second synchronized cardioversion shock may be required to restore normal electrical activity.  If an improper cardioversion shock is delivered, as described above, a patient requires a defibrillation shock.  If the device is still in cardioversion mode when defibrillation is needed, a shock may not be able to be delivered as there are no QRS complexes for the device to synchronize to. These cases can lead to adverse outcomes for the patient, and in rare cases can be fatal if ventricular fibrillation is not terminated. To address this need, we proposed to develop an algorithm to be used in monitor defibrillators, specifically Stryker’s defibrillation devices. We developed two working algorithms as solutions to this problem as well as designed an evaluation protocol to assess the accuracy and reliability of each approach. The first algorithm approach will take in an ECG signal collected from the device and input said signal into a machine learning model that we developed. The model will classify the signal by arrhythmia type and the device would automatically switch into the proper shock mode. The second algorithm approach would detect QRS complexes and R waves in the ECG signals. This would allow the device to deliver a synchronized cardioversion shock timed with the R wave, even if the QRS complex is abnormal. We delivered a “gold standard” database of ECG signals, one algorithm of the proposed two approaches, an algorithm validation protocol, and documentation for implementation. 

Sources


Acknowledgments

Ifrah Javed: LinkedIn here 

Kendall Escene: LinkedIn here

Christopher Neils, Ph.D.: LinkedIn here

Alyssa Taylor, Ph.D.: LinkedIn here

Tyson Taylor, Ph.D.: LinkedIn here

Rob Marx: LinkedIn here

Fred Chapman, Ph.D.: LinkedIn here

Patrick Boyle, Ph.D.: LinkedIn here

Hunter Schafer: Website here

Peter Kudenchuk, MD: Website here