With the accompanying reduction in peak voltages from 4000 V to less than 2000 V, the gas arc relay was dropped; switching is now performed by modern compact insulated-gate bipolar transistors. These are configured in a classic Hâ¿¿bridge circuit, the component that allows motor controllers to run forward or backward [see ”Tender Loving Shock”]. Depending on which two of its four switches are closed, the circuit can deliver a normal or reversed-polarity voltage. Two switches are turned on for 5 to 8 milliseconds to deliver the main shock. Immediately afterward, the two remaining switches are turned on to deliver the residual capacitor charge and perform the burping function of the second phase.
Taken together, these changes reduced the weight of the unit from 40 kg to 1.5 kg and made it safer to operate. Further advances in the capacitor, battery, and high-voltage semiconductors should eventually reduce the size of the AED to that of a deck of playing cards.
The next challenge was to design the brains of the machine. The defibrillator had to figure out, on its own, when to deliver a shock.
The heartbeat is most vulnerable during a period known as the T wave. The T wave occurs just as the heart is beginning to relax after contraction and lasts about 100 ms, or about one-tenth of a heartbeat. A shock administered to a nonfibrillating heart during the T wave could potentially induce fibrillation. If the AED’s electrode patches are positioned far from the target, such as on the belly, the current that finally reaches the heart could be sufficient—if it arrives during that vulnerable period—to induce fibrillation, but it may not be strong enough to then defibrillate and undo its own damage. To provide a check on such situations, AED designs began incorporating an analysis system that checks for a pulse.
Recall that fallen colleague of yours. Once the AED is turned on and you’ve attached the electrode patches, the device’s first task is to recognize the EKG signal to see if ventricular fibrillation has occurred. The system starts by delivering a low-voltage, almost imperceptible 30â¿¿kilohertz signal through the two electrodes. That action measures the impedance to verify a good contact on the body.
When you get an EKG at the hospital or in a clinic, your right leg is used as the ground, or common, electrode from which to measure the tiny voltage differences that make up the EKG. However, as the astute engineer will notice, an AED has only two electrodes—there is no ground electrode. So the average voltage at these two electrodes is used as the reference.
The two-electrode signal is then fed into a very high common-mode-rejection amplifier, which differentiates between the two signals by rejecting the voltages common to both. (Additional complicated circuitry protects this microvolt-sensitive amplifier from the 2000-V shock—20 million times the EKG voltage—delivered to those same electrodes used for the sensing.) A sophisticated peak detector then analyzes the signal in search of a heartbeat. A normal heartbeat is essentially a cycle of blood-pressure increases and decreases, which show up on the cardiogram as clear voltage peaks followed by comparatively flat regions. In ventricular fibrillation, those distinct peaks disappear, and instead a noisy, messy signal will appear on the cardiogram. The peak detector interprets this noisy signal as a series of rapid, randomly spaced heartbeats. The AED makes its initial diagnostic decision by measuring the heart rate. If this rate is more than 150 beats per minute (2.75 hertz), the defibrillation-detection algorithm presumes that ventricular fibrillation has occurred, and the device will announce that the rescuer should administer a shock.
However, dozens of subtle issues can undermine this process. For example, if a patient has an internal pacemaker, that device’s higher-amplitude 60-beats-per-minute pulses could confuse the AED’s peak detector into ignoring the 100â¿¿microvolt ventricular fibrillation signal, so no shock would be administered. The opposite problem can occur if the patient has atrial fibrillation, a fairly benign rhythm disturbance in which the heart still perceivably beats but at irregular intervals. This condition can generate high heart rates, potentially causing the AED to call for an inappropriate shock.
To deal with these confounding rhythms and other interferences, the defibrillation-detection algorithm performs a simple spectral analysis. What follows is one example of such an algorithm; many competing approaches exist. If too much of the signal occurs at a higher frequency (30 to 100 Hz), then noise contamination, perhaps from an ac power line or some skeletal muscle contractions, is suspected and the algorithm will move away from diagnosing ventricular fibrillation. To handle the possibility of atrial fibrillation, the algorithm calculates the average derivative of the EKG voltage. If the average exceeds a critical threshold, that tends to rule out atrial fibrillation. The EKG of a heart in atrial fibrillation has a higher proportion of flat regions of zero voltage, and therefore a zero derivative.
Those tests and more are performed during a three-second window, leading to a tentative diagnosis. The process is then repeated to produce three diagnoses. Only if two or all three analyses indicate ventricular fibrillation will the shock be authorized.
The single-button design was another key improvement over earlier AEDs. An untrained operator who’s facing a panic-filled, life-or-death decision should not be confronted with multiple buttons and expected to quickly figure out which button turns the unit on and which one administers the shock. The breakthrough idea of one of us (Karl) was to have the device turn on automatically and start speaking when its lid is opened.
Every time an AED is used, whether or not a shock is delivered, it sows the seeds for its own improvement. Every EKG is stored in semiconductor memory for later downloading to a computer at the emergency-response headquarters, from which data can be sent back to the AED’s manufacturers. Researchers use this large database of diagnoses to develop and refine future algorithms. Linear methods (such as fast Fourier transforms) and nonlinear techniques (such as neural networks) may soon improve the detection of ventricular fibrillation. These sophisticated signal-processing techniques are being tuned up to make the correct diagnosis, even in the presence of electromagnetic interference, muscle noise, and unusual arrhythmias that might be plaguing the heart. Today’s algorithms may well be supplanted in a few years by more advanced ones that will barely resemble the method described here.
Thanks to the advances in the AED, the weak link in the chain is now CPR. The classic protocol is for a rescuer to administer manual chest compressions and mouth-to-mouth ventilations until someone brings an AED or an ambulance arrives. This keeps blood oxygenated and moving to forestall brain death. Surprisingly, recent studies have shown that the chest compressions also move some air through the lungs, at least for a few minutes after the onset of cardiac arrest. As a result, mouth-to-mouth breathing is now being dropped from those protocols.
However, even trained responders will tire after delivering chest compressions for more than a minute. Researchers at Minneapolis-based Galvani, a company founded by Mark W. Kroll and headed by Gilman, are now exploring an automated electrical form of CPR. Using the same defibrillation patches, this technique relies on complex, lower-voltage waveforms (100 to 200 V) that are delivered once or twice per second and cause strong chest constrictions. The constrictions appear to move blood as effectively as would chest compressions performed by a trained human rescuer.
If this research pans out, in the future a bystander need only attach the patches and the AED will do the rest, performing electrical CPR for a minute or two, followed by a calculated defibrillation shock if it’s needed. Indeed, we may now be on the cusp of a wave of medical automation that allows ordinary individuals to intervene constructively when other people’s lives are at stake. The AED, we think, serves as an important case study for how to fit sophisticated life-saving medical electronics into health care and rehabilitation outside of hospitals. Advances in portable and easy-to-use equipment, home-based therapy, remote health monitoring, and telemedicine may one day allow patients to avoid long, expensive, and emotionally draining hospital visits.
As their efficacy and ubiquity continue to grow, AEDs, we hope, will pave the way for a future where emergency health care is available to all.
