Skip to main content

m22 · Electrogram annotation: local activation time, window of interest, reference, far-field rejection

Warning

Education only. Not for diagnosis, treatment or any clinical decision.

In 60 seconds

An activation map is built from local activation times (LATs). Each LAT is an annotated time on a local electrogram minus the time of a reference electrogram on the same beat. Errors in the reference cancel between points recorded on one beat but add between points recorded on different beats. The annotation rule matters. In the m03 model, the unipolar steepest downstroke falls when the front passes the electrode, and the bipolar peak when it passes the bipole midpoint, as long as the spacing is small enough. An onset rule fires earlier, by an amount that grows with the electrode’s height above the tissue. Noise moves a threshold crossing by noise divided by slope. In a reentry every site activates once per cycle, so the window of interest picks which copy to show; where it wraps, earliest meets latest. Far-field deflections are broad, so their slope-to-amplitude ratio is small. They can be subtracted with a template or blanked by time, at a price.

From the field

  • The window is set from a fixed reference. An early description of magnetic electroanatomic mapping set a window of electrical interest from a stable reference electrogram, usually in the coronary sinus or atrial appendage, wide enough to include atrial activation before P-wave onset (Marchlinski et al. 1998).
  • Window and reference errors spoil reentry maps. Reviewing 43 maps of atrial tachycardia from 14 patients, one group traced poor maps to a window shorter than the cycle length, less than 95% of the cycle mapped, cycle-length variability or an unstable reference, among other causes (Koa-Wing et al. 2015).
  • Far-field signals are common and shaped by anatomy. After pulmonary vein isolation, far-field signals from the superior vena cava appeared in the right superior vein in 35 of 92 patients, with amplitude 0.24 ± 0.11 mV and a sharper main deflection than far-field left atrial signals (Gu et al. 2022).

Derivations

Symbols used throughout. \(t_i\) annotated local time; \(t_{\mathrm{ref}}\) reference time; \(\sigma_a\), \(\sigma_r\) annotation and reference SDs; \(U\) unipolar potential; \(A\) its strength; \(s\) front distance; \(c\) conduction velocity; \(a = w + h\) front width plus electrode height (m03); \(B\) bipolar potential; \(d\) spacing; \(\theta\) angle between the bipole axis and propagation; \(t_{\mathrm{mid}}\) time the front reaches the bipole midpoint; \(f\) onset fraction; \(\sigma_n\) noise SD; \(x\) a signal; \(T\) cycle length; \(w_0\), \(W\) window start and width; \(l\) local signal; \(\phi\) far-field template; \(g\) its gain.

Reference and LAT

\[ \mathrm{LAT}_i = t_i - t_{\mathrm{ref}}, \qquad \operatorname{Var}(\mathrm{LAT}_i - \mathrm{LAT}_j) = 2\sigma_a^2 + \begin{cases} 0 & \text{same beat} \\ 2\sigma_r^2 & \text{different beats} \end{cases} \tag{1}\]

A reference that shifts from beat to beat adds its error to every point collected one beat at a time. In the lab, with an illustrative 0.5-ms annotation error and 3-ms reference error, a plane front fitted to 10 points from one beat had an interquartile range of speed error of 0.04 m/s; spread over 10 beats it was 0.24 m/s (lab part A). Conduction velocity methods are the subject of m25 (Cantwell et al. 2015).

Annotation rules

For the m03 model \(U = As/(s^2 + a^2)\) with \(s = -c(t - t_0)\), and a bipole along the propagation direction:

\[ \frac{dU}{ds} = A\frac{a^2 - s^2}{(s^2 + a^2)^2} \text{ is largest at } s = 0, \qquad B(t_{\mathrm{mid}} + \tau) = B(t_{\mathrm{mid}} - \tau), \quad \lvert B\rvert \text{ peaks at } t_{\mathrm{mid}} \text{ if } d\cos\theta < 2\sqrt3\,a \tag{2}\]

So the unipolar steepest downstroke marks when the front reaches the electrode. This matches the theory and ideal simulations of Steinhaus, who also found differences above 1.8 ms under nonideal conditions (Steinhaus 1989). The bipolar signal is symmetric about the midpoint activation. Its peak sits there while the spacing is below the limit; beyond it the peak splits in two (lab part B). An onset rule fires earlier. In the small-spacing limit:

\[ t_{\mathrm{mid}} - t_{\mathrm{on}} = \frac{a}{c}\sqrt{x_f}, \qquad \frac{1 - x_f}{(1 + x_f)^2} = f \tag{3}\]

For \(f = 0.1\), \(\sqrt{x_f} = 0.84\). With a 2-mm bipole the onset led the midpoint activation by 1.2 ms at 0.5 mm height and 2.9 ms at 3 mm (lab part B). Height changes with contact, so onset LATs carry a bias that varies from point to point.

Exact

Assumptions

Key takeaways

  • Peak and steepest downstroke are unbiased in the ideal model; onset is biased by height.
  • Wide spacing relative to height splits the bipolar peak.

Noise

\[ \sigma_t = \frac{\sigma_n}{\lvert x'(t_c)\rvert} \tag{4}\]

A threshold crossing at \(t_c\) moves by the noise divided by the slope there. In the lab, with noise at 2% of the bipolar peak-to-peak, the onset moved by 0.036 ms as predicted, the peak by 0.11 ms and the unipolar steepest downstroke by 0.20 ms. At 5% the steepest downstroke sometimes locked onto noise, with an SD of 6.4 ms (lab part C). Derivatives amplify fast noise.

Window of interest

\[ \mathrm{LAT}_W = w_0 + \bigl((t - t_{\mathrm{ref}} - w_0) \bmod T\bigr), \quad \text{kept only if } \mathrm{LAT}_W < w_0 + W \tag{5}\]

In a reentry each site activates once per cycle, so its time after the reference is defined only up to whole cycles. The window picks one copy. Where it wraps, the earliest and latest sites sit side by side (“early meets late”), and moving the window moves that line. A window shorter than the cycle leaves part of the circuit out: in the lab a 200-ms window on a 250-ms cycle annotated 80% of the circuit (lab part D). This is why a window shorter than the cycle length spoiled reentry maps (Koa-Wing et al. 2015).

Approximate model

Assumptions

  • Model: an idealized macroreentry with constant speed around the circuit, a stable cycle length and a fixed reference (Equation 5). Tested against epml.annotation.
  • Parameters: cycle length (default 250 ms), window start and width are illustrative and adjustable; the reference sits at a fixed point of the circuit.
  • What it ignores: cycle-length variability, bystander regions and dead-end pathways, slow zones, and annotation error.

Far-field rejection

\[ \frac{\max\lvert dU/dt\rvert}{\operatorname{pp}(U)} = \frac{c}{a}, \qquad x = l + g\phi \;\Rightarrow\; \hat g = \frac{\langle x, \phi\rangle}{\langle \phi, \phi\rangle} = g + \frac{\langle l, \phi\rangle}{\langle \phi, \phi\rangle} \tag{6}\]

Signals from distant tissue are broad: their slope-to-amplitude ratio falls with distance. In the lab a local signal at 1.3 mm had a ratio of 0.77 per ms, an illustrative far-field deflection 0.08 per ms. When the far-field shape is known up to a gain, least squares can subtract it; the gain absorbs any local signal that overlaps the template. A small local signal annotated on the far field before subtraction and correctly after it, while blanking around the far-field time would have discarded a local activation 10 ms away (lab part E). One clinical study cancelled far-field signals from unipolar recordings by least squares, which lowered unipolar voltage by about 25% (Ragot et al. 2021). Simulations show that far-field contributions grow as the electrode moves away from the tissue, and remain even in contact (Martinez-Mateu et al. 2019).

Key takeaways

  • Sharpness separates near from far; overlap in time defeats both blanking and subtraction.
  • A far-field template must not contain the local signal you want.

Clinical correlate

NotePending clinical review

This section describes published findings only. It is not guidance for patient care, and it has not yet been reviewed by an EP physician.

Where this shows up in EP. Every activation map depends on the reference, the window and the annotation rule. Reentry maps need a window close to the cycle length and a stable reference (Koa-Wing et al. 2015). The window is usually anchored to a stable electrogram (Marchlinski et al. 1998). Far-field signals from neighboring structures can resemble local activity (Gu et al. 2022), and removing them changes what a unipolar voltage map shows (Ragot et al. 2021).

What the evidence is. Simulation (Steinhaus 1989; Martinez-Mateu et al. 2019), a retrospective review of tachycardia maps (Koa-Wing et al. 2015), clinical observational studies (Gu et al. 2022; Ragot et al. 2021) and a method review (Cantwell et al. 2015).

What a reader should ask:

Visual summary

Schematic

reference LAT = local − reference onset peak annotation rule window wraps the cycle
Figure 1: LAT from a reference, the annotation rule, and the window of interest.

Assumptions

How experts think

Before trusting an activation map, ask:

  1. Is the reference stable from beat to beat? (Equation 1)
  2. Which rule annotated the points, and does it depend on contact? (Equation 2, Equation 3)
  3. How noisy were the signals, and how steep at the annotation? (Equation 4)
  4. Does the window cover the whole cycle, and where does it wrap? (Equation 5)
  5. Could any annotation be on a far-field signal? (Equation 6)

One sentence to keep: a LAT is a choice of reference, window and rule, not a measurement of the heart alone.

Cheat sheet

Quantity Result
LAT (Equation 1) \(t_i - t_{\mathrm{ref}}\); difference variance \(2\sigma_a^2\) in one beat, \(+2\sigma_r^2\) across beats
Rules (Equation 2) steepest downstroke at \(s = 0\); bipolar peak at midpoint if \(d\cos\theta < 2\sqrt3\,a\)
Onset (Equation 3) leads by \((a/c)\sqrt{x_f}\); \((1 - x_f)/(1 + x_f)^2 = f\)
Noise (Equation 4) \(\sigma_t = \sigma_n/\lvert x'\rvert\)
Window (Equation 5) \(w_0 + ((t - w_0) \bmod T)\), kept if \(< w_0 + W\)
Far field (Equation 6) sharpness \(c/a\); \(\hat g = g + \langle l, \phi\rangle/\langle\phi, \phi\rangle\)

Values used in this module

Value Source
Window from a fixed CS or appendage reference Marchlinski et al. (1998)
43 maps, 14 patients; window < cycle, < 95% mapped, unstable reference among causes of poor maps Koa-Wing et al. (2015)
Far-field SVC signals in 35/92 patients; 0.24 ± 0.11 mV Gu et al. (2022)
Unipolar voltage about 25% lower after far-field cancellation; 12 patients Ragot et al. (2021)
Activation-time differences above 1.8 ms under nonideal conditions (simulation) Steinhaus (1989)
Conduction velocity 1.0 m/s; height 1 mm; spacing 2 mm Heil et al. (2023), Riccio et al. (2021), Intracardiac Atrial Fibrillation Database (Version 1.0.0) (2003)
Noise levels, reference jitter, cycle length, far-field shape illustrative
Onset leads, jitters, window coverage, gains our own code

Interview questions

Try each question before you open its model answer. In a real interview, say which assumptions you are making.

Q1. Define local activation time.

Model answer

The annotated time on the local electrogram minus the annotated time on a reference electrogram in the same beat (Equation 1).

Q2. Why does a multielectrode catheter give better local conduction-velocity estimates than single points from different beats?

Model answer

Points from one beat share the reference error, which cancels in their differences; points from different beats each carry their own (Equation 1).

Q3. Why is the unipolar steepest downstroke used as the activation time?

Model answer

In theory and ideal simulation it coincides with the fastest upstroke of the local action potential, the moment the front passes (Equation 2) (Steinhaus 1989).

Q4. When does the bipolar peak mark the midpoint activation, and when not?

Model answer

In the m03 model, when \(d\cos\theta < 2\sqrt3\,a\); with wider spacing relative to height the peak splits into two (Equation 2).

Q5. What bias does an onset rule introduce?

Model answer

It fires before the front arrives, by \((a/c)\sqrt{x_f}\) in the small-spacing limit, so the bias grows with electrode height and falls with conduction velocity (Equation 3).

Q6. How does noise affect a threshold-crossing annotation?

Model answer

It moves it by noise SD over the slope at the crossing (Equation 4).

Q7. Why can the steepest-downstroke rule fail badly on noisy signals?

Model answer

It uses a derivative, which amplifies fast noise; a noise spike can have a steeper slope than the true downstroke (Equation 4).

Q8. What does the window of interest do in a reentry?

Model answer

It chooses which cycle’s copy of each activation to show, by wrapping times modulo the cycle length into the window (Equation 5).

Q9. What happens if the window is shorter than the cycle length?

Model answer

Part of the circuit is not annotated or is assigned to the wrong cycle (Equation 5); this was among the causes of poor reentry maps in one review (Koa-Wing et al. 2015).

Q10. How can you tell a far-field deflection from a local one?

Model answer

Far-field signals are broad, with a small slope-to-amplitude ratio (Equation 6); timing relative to the far-field source is another clue.

Q11. What is the risk of subtracting a far-field template?

Model answer

If local activity overlaps the template, the fitted gain absorbs part of it (Equation 6) (Ragot et al. 2021).

Q12. What is the risk of blanking?

Model answer

Any local activation inside the blanked span is lost (Equation 6).

Code lab

The lab annotates synthetic electrograms.

Part What you do Data
A The reference and its error (Equation 1) synthetic
B Annotation rules (Equation 2, Equation 3) synthetic
C Noise moves annotations (Equation 4) synthetic
D The window of interest (Equation 5) synthetic
E Far-field signals (Equation 6) synthetic

Each part ends with checks: lines that pass with the code as given. If you change something and a check fails, that is information, not an error to hide.

Self-check

Short checks you can do in your head. Open each answer after you have tried it.

S1. Annotation SD 1 ms, reference SD 2 ms. SD of a LAT difference across two beats?

Answer

\(\sqrt{2 + 8} \approx\) 3.2 ms (Equation 1).

S2. The same two points recorded on one beat?

Answer

\(\sqrt2 \approx\) 1.4 ms (Equation 1).

S3. \(a\) = 1 mm. Largest bipolar spacing (along the propagation direction) that keeps the peak at the midpoint?

Answer

\(2\sqrt3 \approx\) 3.5 mm (Equation 2).

S4. Noise 0.02 mV, slope 0.5 mV/ms at the crossing. Timing SD?

Answer

0.04 ms (Equation 4).

S5. Cycle 300 ms, window −200 to 100 ms. A site activates 150 ms after the reference. Assigned LAT?

Answer

−150 ms (150 − 300) (Equation 5).

S6. Speed 1 mm/ms, \(a\) = 2 mm. Slope-to-amplitude ratio of the unipolar signal?

Answer

0.5 per ms (Equation 6).

Literature figures

These figures are worth looking at alongside this module. We link to them instead of copying them: some licenses do not allow reuse here, and others we could not confirm. Each entry says what to look for and where this module explains it.

  • Activation times from electrograms. Steinhaus’s comparisons of electrogram and action-potential timing in simulated cables (Steinhaus 1989). This module: annotation rules.
  • Map grading and its pitfalls. Koa-Wing and colleagues’ examples of poor maps from window and reference settings (Koa-Wing et al. 2015). This module: window of interest.
  • Far-field cancellation. Ragot and colleagues’ unipolar recordings before and after cancellation (Ragot et al. 2021). This module: far-field rejection.

Further reading

  • Methods. A review of LAT annotation and conduction-velocity algorithms (Cantwell et al. 2015).
  • Reading maps. A tutorial review of what activation-map colors mean (Del Carpio Munoz et al. 2010).
  • Far field. Simulation of far-field contributions (Martinez-Mateu et al. 2019); clinical far-field signals (Gu et al. 2022).
  • Related modules. m03 (electrograms), m04 (recording chain), m11 (EP study), m24 (activation maps), m25 (conduction velocity), m26 (map uncertainty).

References

Cantwell, C D, C H Roney, F S Ng, et al. 2015. “Techniques for Automated Local Activation Time Annotation and Conduction Velocity Estimation in Cardiac Mapping.” Computers in Biology and Medicine 65: 229–42. https://doi.org/10.1016/j.compbiomed.2015.04.027.
Del Carpio Munoz, Freddy, Traci L. Buescher, and Samuel J. Asirvatham. 2010. “Three-Dimensional Mapping of Cardiac Arrhythmias: What Do the Colors Really Mean?” Circulation: Arrhythmia and Electrophysiology 3 (6): e6–11. https://doi.org/10.1161/CIRCEP.110.960161.
Gu, Wentao, Weizhuo Liu, Jian Li, et al. 2022. “Anatomy-Based Characteristics of Far-Field SVC Electrograms in Right Superior Pulmonary Veins After Isolation.” Scandinavian Cardiovascular Journal 56 (1): 224–30. https://doi.org/10.1080/14017431.2022.2095015.
Heil, Emanuel, Jin-Hong Gerds-Li, Julian Keznickl-Pulst, Vesna Furundzija-Cabraja, et al. 2023. “Left Atrial Conduction Times and Regional Velocities in Persistent Atrial Fibrillation Patients with and Without Fibrotic Atrial Cardiomyopathy.” Heart and Vessels 38 (10): 1277–87. https://doi.org/10.1007/s00380-023-02282-5.
Intracardiac Atrial Fibrillation Database (Version 1.0.0). 2003. PhysioNet. https://doi.org/10.13026/C23S33.
Koa-Wing, Michael, Hiroshi Nakagawa, Vishal Luther, et al. 2015. “A Diagnostic Algorithm to Optimize Data Collection and Interpretation of Ripple Maps in Atrial Tachycardias.” International Journal of Cardiology 199: 391–400. https://doi.org/10.1016/j.ijcard.2015.07.017.
Marchlinski, F., D. Callans, C. Gottlieb, E. Rodriguez, R. Coyne, and D. Kleinman. 1998. “Magnetic Electroanatomical Mapping for Ablation of Focal Atrial Tachycardias.” Pacing and Clinical Electrophysiology 21 (8): 1621–35. https://doi.org/10.1111/j.1540-8159.1998.tb00252.x.
Martinez-Mateu, Laura, Lucia Romero, Javier Saiz, and Omer Berenfeld. 2019. “Far-Field Contributions in Multi-Electrodes Atrial Recordings Blur Distinction Between Anatomical and Functional Reentries and May Cause Imaginary Phase Singularities: A Computational Study.” Computers in Biology and Medicine 108: 276–87. https://doi.org/10.1016/j.compbiomed.2019.02.022.
Ragot, Don, Sachin Nayyar, Sophia Z. Massin, et al. 2021. “Unipolar Electrogram-Based Voltage Mapping with Far-Field Cancellation to Improve Detection of Abnormal Atrial Substrate During Atrial Fibrillation.” Journal of Cardiovascular Electrophysiology 32 (6): 1572–83. https://doi.org/10.1111/jce.14999.
Riccio, Jennifer, Alejandro Alcaine, Sara Rocher, Laura Martinez-Mateu, et al. 2021. “Characterization of Atrial Propagation Patterns and Fibrotic Substrate with a Modified Omnipolar Electrogram Strategy in Multi-Electrode Arrays.” Frontiers in Physiology 12: 674223. https://doi.org/10.3389/fphys.2021.674223.
Steinhaus, B. M. 1989. “Estimating Cardiac Transmembrane Activation and Recovery Times from Unipolar and Bipolar Extracellular Electrograms: A Simulation Study.” Circulation Research 64 (3): 449–62. https://doi.org/10.1161/01.res.64.3.449.