Arrhythmia detection algorithms and shock/therapy delivery by an implantable cardioverter-defibrillator
An ICD never sleeps. From the moment it is implanted, the device's sense amplifier continuously listens to the intracardiac electrogram (EGM) picked up by the right-ventricular lead — a far more direct and noise-resistant signal than a surface ECG — timing every ventricular depolarization and comparing its rate and shape to a stored baseline.
Unlike a 12-lead surface ECG, the ICD reads a bipolar near-field electrogram between two electrodes just millimeters apart on the RV lead (tip and ring, or coil). This geometry sharply attenuates far-field signals from atrial activity, skeletal muscle, and electromagnetic noise, producing a clean, high-amplitude local ventricular signal ideal for automated rate counting.
A sense amplifier with auto-adjusting gain and a blanking/refractory period after each detected event prevents the device from double-counting a single wide QRS complex or reacting to T-wave amplitude.
The ICD does not think in "beats per minute" — it thinks in milliseconds between consecutive sensed events (RR intervals). Every sensed ventricular event starts a new timer; the interval to the next sensed event is classified into a rate bin instantly.
A 60 bpm rhythm produces ~1000 ms intervals; 200 bpm produces 300 ms intervals. This interval-based logic is what allows extremely fast, deterministic zone classification without any need to average heart rate over a full minute.
During confirmed normal sinus rhythm, the device automatically captures and stores a template of the typical QRS shape — amplitude, width, and vector — sampled from the far-field shock-coil-to-can electrogram. This template becomes the reference used later by morphology discriminators to judge whether a fast beat still "looks like" the patient's normal conducted rhythm or looks abnormal, as ventricular-origin beats typically do.
When the ventricular rate climbs, the ICD does not react to a single fast beat. It evaluates the rate against physician-programmed rate cutoffs that define discrete detection zones — commonly VT, sometimes an intermediate Fast-VT (FVT), and VF — each requiring its own count of qualifying short intervals before therapy is armed.
Each zone boundary is a clinical decision. Set the VT cutoff too low (too sensitive) and sinus tachycardia during exercise may trigger unnecessary detection; set it too high (too specific) and true slow VT may go undetected until it degenerates further. Contemporary programming favors higher cutoff rates and longer detection durations specifically to reduce inappropriate therapy for non-lethal fast rhythms, informed by trials such as MADIT-RIT and PROVIDE.
Detection is not simply "rate above cutoff" — it requires a run of qualifying intervals. A typical VF algorithm requires 8 of the last 10 sensed intervals to be shorter than the VF cutoff interval; a VT algorithm might require 16 of 24. This counting scheme tolerates a few missed or oversensed beats without either delaying true detection or triggering on isolated ectopic beats.
Once the count threshold is reached, the arrhythmia is "detected" and the device proceeds toward discrimination and therapy — but detection itself is reversible: if intervals lengthen back into the normal range, the counter resets.
Even before formal discrimination, the device notes how abruptly the rate accelerated. A gradual rise over many beats (typical of sinus tachycardia from exercise, fever, or anxiety) is treated very differently from an abrupt jump from a sinus interval directly into the VT/VF interval range — a hallmark of a true reentrant ventricular arrhythmia beginning with a premature ventricular contraction.
A fast rate alone is not proof of a lethal ventricular arrhythmia — rapidly conducted atrial fibrillation, atrial flutter, or sinus tachycardia can all enter the VT zone. Before committing a patient to a painful shock, modern ICDs run a discrimination cascade combining onset, stability, and morphology analysis, and in dual-chamber devices, the atrial-to-ventricular relationship.
The onset discriminator compares the interval immediately preceding tachycardia to the intervals during tachycardia. Sinus tachycardia accelerates gradually beat-by-beat as catecholamine drive rises; VT classically begins abruptly, often with a single premature beat dropping the interval sharply. An abrupt drop favors VT; a gradual ramp favors sinus tachycardia and withholds therapy.
The stability discriminator measures beat-to-beat variability in the RR interval during the fast rhythm. Monomorphic VT is characteristically very regular (low variability); atrial fibrillation conducted rapidly to the ventricle is characteristically irregularly irregular (high variability). Intervals that vary by more than the programmed stability threshold bias the classification toward AF rather than VT.
The morphology (wavelet) discriminator compares the shape of the far-field electrogram during each fast beat to the sinus-rhythm template captured in Stage 1. A supraventricular beat conducted normally through the His-Purkinje system usually retains a QRS shape similar to baseline (high correlation). A ventricular-origin beat, arising outside the normal conduction system, typically produces a markedly different, wider shape (low correlation).
Dual-chamber devices add one more powerful clue: the AV relationship. If atrial activity is faster than or dissociated from ventricular activity, that strongly favors an atrial arrhythmia; 1:1 AV association with abrupt onset strongly favors VT.
Once a ventricular arrhythmia is confirmed, the ICD escalates therapy in a programmed hierarchy: painless antitachycardia pacing (ATP) is attempted first for monomorphic VT in the slower zones, and if it fails — or the rhythm is VF or too fast for pacing to catch — the device charges its capacitors and delivers a synchronized or unsynchronized biphasic high-voltage shock.
Monomorphic VT is usually sustained by a reentrant electrical circuit looping through scarred myocardium. ATP delivers a short burst of pacing pulses faster than the tachycardia (typically 81–88% of the VT cycle length), aiming to "catch" the excitable gap in the reentrant loop and collide with — and extinguish — the circulating wavefront. Because it uses low-voltage pacing pulses rather than a shock, ATP is imperceptible or only mildly noticeable to the patient, and successfully terminates roughly 90% of slower monomorphic VTs, sparing the patient a painful shock.
If ATP fails, is not attempted (e.g., in VF or very fast VT), or the rhythm accelerates during pacing, the device begins charging its high-voltage capacitors from the battery — a process that takes several seconds and is itself a measurable clinical parameter (charge time). The stored energy is delivered as a biphasic truncated exponential waveform: current flows in one polarity for several milliseconds, then reverses polarity for a second, shorter phase. Biphasic waveforms lower the defibrillation threshold substantially compared to older monophasic designs, allowing effective termination with less delivered energy and less myocardial damage.
When possible (organized VT with a discernible R-wave), the shock is synchronized to the sensed R-wave to avoid delivering current during the vulnerable T-wave period, which could itself induce VF (the "R-on-T" phenomenon). True VF, being disorganized with no reliable R-wave to sync to, is treated with an unsynchronized shock delivered as soon as charging completes.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Antitachycardia Pacing (ATP) | Slow, monomorphic VT | Burst/ramp overdrive pacing collides with reentrant wavefront | Painless, no charge time, battery-sparing |
| Low-energy cardioversion | Persistent VT after failed ATP | Synchronized shock at lower joules than max output | Terminates while limiting energy delivered |
| Maximum-output defibrillation | VF / polymorphic VT | Unsynchronized biphasic shock at full programmed energy | Highest probability of first-shock success |
Therapy delivery is not the end of the episode. The device immediately re-examines the rhythm to confirm success, logs the full episode electrogram for the clinic to review, and — because every high-voltage charge draws meaningfully on a finite battery — quietly tracks how much longevity that shock just consumed.
Immediately after a shock, the device re-senses the rhythm through a brief blanking period and then re-evaluates rate and regularity. If sinus rhythm is confirmed, the episode closes and is logged for clinician review at the next interrogation. If the tachyarrhythmia persists, the device redetects and escalates — typically to progressively higher-energy shocks — up to the programmed maximum number of therapy attempts per episode.
Electrolytic high-voltage capacitors gradually lose their ability to hold charge efficiently if left idle, which would otherwise lengthen charge time when a real shock is eventually needed. To prevent this, the device automatically performs a brief "reformation" charge-and-dump cycle roughly every three months — invisible to the patient — to keep charge time consistently fast. Charge time itself is monitored at every follow-up; a significant prolongation can signal capacitor degradation or approaching battery depletion, prompting the Elective Replacement Indicator (ERI) and eventually End of Life (EOL) alerts.
ICD batteries are engineered to deliver years of continuous low-current sensing and occasional pacing on milliamp-scale currents — but a single maximum-energy shock draws a comparatively enormous current pulse to charge the capacitors. Frequent shocks, whether appropriate (treating real VT/VF) or inappropriate (triggered by SVT misclassification, lead noise, or oversensing), measurably shorten device longevity and hasten the need for generator replacement. This is a central reason discrimination algorithms and conservative zone programming matter clinically: minimizing unnecessary shocks preserves both patient wellbeing and battery life.