Skip to main content

m03 · Unipolar, bipolar and omnipolar electrograms as spatial filters

Warning

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

Note

Pilot module, draft. All six stages are written. Every equation and every number from our own code is checked by tests in tests/test_m03_*.py, and every source was looked up. Nothing here has had expert review yet, and the clinical correlate awaits an EP physician.

In 60 seconds

An electrogram is the voltage an electrode picks up from currents crossing cell membranes nearby. A unipolar electrogram measures one electrode against a distant reference. It sees the local wavefront clearly, but it also sees activity far away. A bipolar electrogram subtracts two nearby unipolar signals. That subtraction acts like a spatial derivative: it keeps sharp local detail and cancels slowly varying, distant activity. The price is direction dependence. A bipole aligned with the wave records a large signal; a bipole turned across it records almost nothing. An omnipolar signal combines several bipoles to estimate the local electric field, then reads it along the direction the wave is travelling, which removes most of that dependence. Height above the tissue, electrode size and electrode spacing all act as spatial filters on what the catheter can see.

From the field

Practical lessons reported from mapping labs, each tied to the part of this module that explains it.

  • Use unipolar and bipolar recordings together. A review of recording techniques describes minimally filtered unipolar recordings as useful for mapping a focal source (a small region from which activation spreads outward), but notes that their far-field signal can hide small local signals in diseased tissue. Bipolar recordings are standard because they reject far-field signal. Still, the local signal can come from under either electrode, far-field signals do appear, and unipolar recordings taken at the same time help in some situations (Stevenson and Soejima 2005). The physics is in Equation 13.
  • Filtering changes what a unipolar signal can tell you. High-pass filtering a unipolar recording reduces its far-field part but limits what its shape can tell you (Stevenson and Soejima 2005). Module m04 covers the recording chain and its filters.
  • Catheter orientation changes bipolar voltage. In sheep with a myocardial infarction (heart muscle killed by loss of its blood supply), a grid catheter compared two bipoles at each site that shared the same center and spacing but crossed at right angles. Their voltages differed by a median of 31.7%, and by 48.7% where the wavefront ran perpendicular to one bipole and parallel to the other. Sites between 0.5 and 1.5 mV were the most likely to change voltage range: only 55.8% stayed in the same range (Takigawa et al. 2018). Demo m03-d2 and Equation 10 show why.
  • Smaller, closer electrodes see more inside low-voltage areas. A low-voltage area is a region whose bipolar voltage falls below a chosen cut-off; it is often read as scar (fibrous tissue that has replaced heart muscle). In patients with scar-related arrhythmias (abnormal heart rhythms) of the atria (the upper chambers of the heart), a catheter with 1-mm electrodes measured a smaller area below 0.5 mV than one with a 3.5-mm tip (14.7 versus 20.4 cm²) and higher voltages inside it (0.28 versus 0.17 mV). It also found distinct, timeable electrograms at 54.4% of low-voltage points, against 21.4% for the larger tip (Anter et al. 2015). The catheters differed in both electrode size and spacing, so this is their combined effect; Equation 11 and demo m03-d3 show both filters.
  • Omnipoles reduce orientation effects in real maps. In isolated rabbit, pig and human hearts, bipolar voltage maps made along two perpendicular electrode orientations differed by an average of 0.25 ± 0.18 mV in the human hearts. Omnipolar voltages were steadier from beat to beat (coefficient of variation 0.08 versus 0.28 for bipoles, in human hearts) (Magtibay et al. 2017). See Equation 15 and demo m03-d5.

Derivations

Symbols used throughout. \(\sigma\) extracellular conductivity (S/m). \(i_m\) membrane current per unit area of tissue, positive outward (A/m²). \(V_m\) transmembrane voltage. \(h\) electrode height above the tissue (mm). \(w\) width of the activation front (mm). \(c\) conduction velocity (m/s, which equals mm/ms). \(\hat n\) unit vector along the direction of propagation. \(d\) interelectrode spacing (mm). \(\hat u\) unit vector along a bipole, from its proximal to its distal electrode. \(\theta\) angle between \(\hat u\) and \(\hat n\). \(k\) spatial frequency (rad/mm).

Where a unipolar electrogram comes from

The extracellular electrogram is produced by transmembrane currents that flow out into the extracellular space as the impulse propagates (Bakker 2019). A unipolar recording measures the potential at one exploring electrode relative to a distant reference electrode.

To compute that potential, treat the tissue and blood as an infinite, homogeneous, isotropic conductor, and treat the fields as quasi-static. A membrane current density then produces a potential that falls off as one over distance (Plonsey and Barr 2007). For a thin sheet of tissue in the plane \(z = 0\):

\[ \phi(\mathbf r, t) = \frac{1}{4\pi\sigma} \iint \frac{i_m(\mathbf r', t)}{\lVert \mathbf r - \mathbf r' \rVert}\, dA' \tag{1}\]

Membrane current from the voltage profile. Along a thin strand, the intracellular current is \(I_i = -\beta\,\partial V_m/\partial x\), where \(\beta\) (S) lumps the intracellular conductivity and the sheet thickness. If the extracellular potential gradient is small next to the intracellular one, current conservation says the outward membrane current equals the loss of intracellular current along the strand, \(i_m = -\partial I_i/\partial x\), so

\[ i_m = \beta\,\frac{\partial^2 V_m}{\partial x^2} \tag{2}\]

A wavefront with a closed form. Let a planar front travel along \(+x\) at speed \(c\). Write \(s = x - ct\) for the signed distance from the front: \(s > 0\) is tissue the front has not reached yet. Model the rise of \(V_m\) across the front with an arctangent profile of width \(w\) and amplitude \(\Delta V\):

\[ V_m(s) = \Delta V \left( \tfrac12 - \tfrac1\pi \arctan\frac{s}{w} \right), \qquad \frac{dV_m}{ds} = -\frac{\Delta V}{\pi}\,\frac{w}{s^2 + w^2} \tag{3}\]

The slope is a Lorentzian of width \(w\). Its second derivative is odd in \(s\): outward current (a source) ahead of the front, inward current (a sink) behind it. Net membrane current over the whole front is zero.

The height filter. Take the Fourier transform of Equation 1 in the plane of the sheet. The kernel \(1/\sqrt{\rho^2 + h^2}\) transforms to \(2\pi e^{-|k| h}/|k|\), so

\[ \hat\phi(k) = \frac{e^{-|k| h}}{2\sigma |k|}\, \hat i_m(k) \tag{4}\]

The factor \(e^{-|k|h}\) is a spatial low-pass filter. The farther the electrode sits from the tissue, the more fine spatial detail it loses.

The unipolar electrogram. Insert Equation 2 and Equation 3 into Equation 4. For a planar front only the spatial frequency along \(x\) matters, so \(k\) is one-dimensional. Since \(\hat i_m = -\beta k^2 \hat V_m\) and the transform of \(dV_m/ds\) is \(-\Delta V e^{-w|k|}\), the potential spectrum is \(\hat\phi = -i\,\mathrm{sgn}(k)\,\frac{\beta\Delta V}{2\sigma}\, e^{-(w+h)|k|}\). Its inverse transform is

\[ U(s) = A\,\frac{s}{s^2 + a^2}, \qquad A = \frac{\beta\,\Delta V}{2\pi\sigma}, \qquad a = w + h \tag{5}\]

Read it as a waveform in time at a fixed electrode, where \(s = x_e - ct\):

  • Before the front arrives (\(s > 0\)) the potential is positive. It peaks at \(s = a\).
  • It falls through zero exactly as the front passes under the electrode (\(s = 0\)).
  • After the front passes (\(s < 0\)) it is negative, with its trough at \(s = -a\).
  • Peak-to-peak amplitude is \(A/a = A/(w + h)\). Moving the electrode away from the tissue, or widening the front, shrinks the signal.

Slope and the intrinsic deflection. By the chain rule,

\[ \frac{dU}{dt} = -c\,\frac{dU}{ds} \tag{6}\]

so the unipolar slope scales with conduction velocity. In this model the steepest negative slope is \(-cA/a^2\), and it occurs exactly at \(s = 0\), when the front passes the electrode. That is the basis of the classic marker of local activation time, the moment the front reaches the electrode:

\[ t_{\text{act}} \approx \arg\max_t \left( -\frac{dU}{dt} \right) \tag{7}\]

Simulation work predicts that this “intrinsic deflection” matches the steepest rise of the action potential, with small errors under ideal uniform propagation. It errs more near changes in activation sequence, short strands and non-uniform coupling (Steinhaus 1989). Our model is the ideal case, so the match is exact.

Approximate model

Assumptions

  • Model: the arctangent front of Equation 3 in a thin tissue sheet, seen by a point electrode in an infinite homogeneous conductor (Equation 5). It ignores the blood–tissue boundary, anisotropy, curved fronts and amplifier filtering.
  • Default conduction velocity 1.0 m/s, reported for the anterior left atrial wall in normal-voltage areas; low-voltage areas in the same study averaged 0.6 m/s (Heil et al. 2023).
  • Default electrode height 1 mm, the electrode-to-tissue distance used in a published omnipolar simulation (Riccio et al. 2021).
  • Default front width 0.3 mm. From Equation 3, the fastest upstroke is \(c\,\Delta V/(\pi w)\), so \(w = c\,\Delta V/(\pi \max dV_m/dt)\). Human right atrial trabeculae had an action-potential amplitude \(\Delta V\) of about 95–99 mV (Loose et al. 2014). With an upstroke of about 100 V/s in human atrial fibres (Wang et al. 1991) and \(c\) = 1 m/s, \(w \approx\) 0.3 mm. The trabeculae study itself reported 250–300 V/s, which gives \(w \approx\) 0.1 mm. Because \(h\) = 1 mm dominates \(a = w + h\), that range changes \(a\) by only 0.2 mm; set the front-width slider to 0.1 mm to compare (arithmetic checked in tests/test_m03_unipolar.py).
  • Amplitudes are relative. The scale \(A\) depends on tissue properties not modeled here, so the readout compares with the defaults.
  • What would change the result: non-ideal propagation moves the steepest downstroke away from true activation (Steinhaus 1989).

Key takeaways

  • A unipolar electrogram is a weighted sum of membrane currents, weighted by one over distance.
  • For a passing planar front it is positive, then crosses zero at activation, then is negative.
  • Its amplitude falls as \(1/(w + h)\); its steepest slope scales with \(c/(w + h)^2\).
  • The steepest downstroke marks activation exactly only in ideal propagation.

The bipolar electrogram as a spatial difference

A bipolar electrogram is the difference between two unipolar signals. We use the convention distal minus proximal:

\[ V_{\text{bip}}(t) = \phi(\mathbf r_1, t) - \phi(\mathbf r_2, t), \qquad \mathbf r_{1,2} = \mathbf r_0 \pm \tfrac{d}{2}\,\hat u \tag{8}\]

For a planar wave, only the distance of each electrode along \(\hat n\) matters. The two electrodes sit at \(s \pm \delta\) with \(\delta = \tfrac{d}{2}\cos\theta\), so

\[ V_{\text{bip}}(s) = U(s + \delta) - U(s - \delta) \tag{9}\]

A close bipole is a directional derivative. When \(d\) is small next to \(a\), expand to first order:

\[ V_{\text{bip}} \approx d\, \hat u \cdot \nabla \phi = d \cos\theta\, \frac{dU}{ds} \tag{10}\]

The relative error of this approximation falls as \(d^2\). Three consequences follow:

  • Direction dependence. Amplitude scales with \(|\cos\theta|\). A bipole perpendicular to the direction of propagation (\(\theta = 90°\)) records nothing, because both electrodes activate together. In animal recordings, voltage at the same site varied most where the wavefront was perpendicular to one bipole and parallel to the other (Takigawa et al. 2018).
  • Amplitude. Using \(dU/ds\) from Equation 5, the peak-to-peak bipolar amplitude tends to \(d\,|\cos\theta| \cdot 9A/(8a^2)\) as \(d \to 0\).
  • Locality. Unipolar amplitude falls as \(1/a\), but bipolar amplitude falls as \(1/a^2\). The bipole is dominated by tissue closer to the catheter.

Approximate model

Assumptions

  • Model: two point electrodes at \(\pm d/2\) along the bipole axis and the planar arctangent front (Equation 9). It ignores electrode size (see m03-d3), curved fronts and the blood–tissue boundary.
  • Default spacing 2 mm, the bipole spacing of a 2-5-2 mm decapolar (ten-electrode) catheter (Intracardiac Atrial Fibrillation Database (Version 1.0.0) 2003). Height 1 mm (Riccio et al. 2021) and conduction velocity 1.0 m/s (Heil et al. 2023) as in m03-d1; front width as in m03-d1.
  • The polar plot divides by an aligned bipole of the same spacing, so it shows direction dependence only.
  • What would change the result: curved or colliding wavefronts, and slow conduction zones under one electrode.

Key takeaways

  • A bipole measures a spatial difference; when it is short, that is a directional derivative.
  • Bipolar amplitude depends on catheter orientation and on spacing, not only on the tissue.
  • A bipole is more local than a unipole: its amplitude falls off faster with distance.

Recordings as spatial filters

Equation 4 already showed that distance from the tissue is a low-pass filter. Two more filters complete the picture.

Electrode size. A disk electrode of radius \(r_e\) reports, approximately, the average potential over its face. Averaging a spatial harmonic over a disk multiplies it by

\[ H_{\text{disk}}(k) = \frac{2 J_1(k r_e)}{k r_e} \tag{11}\]

where \(J_1\) is the Bessel function of the first kind. This is another low-pass filter: larger electrodes blur fine detail. The model ignores contact effects and current shunting by the metal.

The bipole. Apply Equation 8 to a harmonic \(e^{i k\, \hat n \cdot \mathbf r}\):

\[ H_{\text{bip}}(k) = 2i \sin\!\left( \frac{k d \cos\theta}{2} \right), \qquad |H_{\text{bip}}| = 2 \left| \sin\!\left( \frac{k d \cos\theta}{2} \right) \right| \tag{12}\]

This is a spatial high-pass filter. It rejects slowly varying fields (\(k \to 0\)), peaks near \(k d \cos\theta = \pi\), and has nulls wherever \(k d \cos\theta\) is a multiple of \(2\pi\).

Put together, a bipolar recording sees the tissue’s current pattern through \(H_{\text{bip}} \cdot H_{\text{disk}} \cdot e^{-|k|h}\): a spatial band-pass filter. Height and electrode size set the upper edge; spacing sets the lower edge and the nulls.

Far-field rejection. A distant source changes slowly across the catheter, so the bipole mostly cancels it. For a distant point source the unipole falls off as \(1/R\) and a bipole as \(1/R^2\); for a distant dipole, as \(1/R^2\) and \(1/R^3\):

\[ \phi_{\text{monopole}} \propto \frac{1}{R},\quad V_{\text{bip}} \propto \frac{d}{R^2}; \qquad \phi_{\text{dipole}} \propto \frac{1}{R^2},\quad V_{\text{bip}} \propto \frac{d}{R^3} \tag{13}\]

This is why bipolar recordings reject far-field signal better than unipolar ones. They do not remove it completely: far-field signals still appear in bipolar recordings, and the local signal can lie under either electrode (Stevenson and Soejima 2005).

Exact

Assumptions

Approximate model

Assumptions

  • Local signal: as in m03-d2, with the bipole aligned with the wave and 2-mm spacing (Intracardiac Atrial Fibrillation Database (Version 1.0.0) 2003).
  • Far source: one current dipole on the bipole axis, pointing along it, with a biphasic time course 25 ms after local activation. Its strength is set so that at 20 mm it equals the local unipolar signal. These are illustrative choices, not physiological values; only the trend with distance is meant (Equation 13).
  • Each panel is scaled by its own local peak-to-peak, so the panels compare the far-field share, not absolute size.
  • What would change the result: a far source off the bipole axis or a bipole across it (less bipolar pickup), and conductivity boundaries such as the blood–tissue interface.

Key takeaways

  • Distance and electrode size are spatial low-pass filters; a bipole is a spatial high-pass filter.
  • A bipolar recording is therefore a spatial band-pass filter, tuned by height, size and spacing.
  • Bipoles reject distant sources better than unipoles, but not completely.

Omnipolar signals from a clique of electrodes

An omnipolar method estimates the local electric field from a small group of electrodes, called a clique, and reads it along the direction of propagation (Deno et al. 2017; Riccio et al. 2021).

The field from a clique. The electric field is \(\mathbf E = -\nabla\phi\). If the potential is nearly linear across the clique, each bipole between electrodes \(i\) and \(j\) is a projection of the field, \(b_{ij} = \phi_j - \phi_i \approx -(\mathbf r_j - \mathbf r_i) \cdot \mathbf E\). Stack the bipoles into a vector \(\mathbf b\) and the electrode separations into the columns of a matrix \(D\). Then \(\mathbf b = -D^{\mathsf T}\mathbf E\), and the least-squares estimate is (Riccio et al. 2021)

\[ \hat{\mathbf E}(t) = -\left( D D^{\mathsf T} \right)^{-1} D\, \mathbf b(t) \tag{14}\]

A triangle of three electrodes or a square of four can be used. For a linear potential the estimate is exact.

The omnipolar signal. For a plane wave the field points along \(\pm\hat n\) at every instant, so \(\hat{\mathbf E}(t)\) traces a line, not a loop, along the propagation direction (Riccio et al. 2021). Reading the field along a unit direction \(\hat\omega\) gives the signal a bipole of spacing \(d\) would record if it pointed that way:

\[ v_\omega(t) = -d\, \hat\omega \cdot \hat{\mathbf E}(t) \tag{15}\]

With \(\hat\omega = \hat n\), this equals an aligned bipole, whatever the catheter’s orientation. Here we take \(\hat\omega\) as the principal axis of the field loop. The published method instead picks the angle whose field projection best matches the time derivative of the unipolar signal (Riccio et al. 2021), following Deno et al. (2017). For a plane wave both choices give \(\hat n\).

Velocity from one clique. Along \(\hat n\) the field is \(-dU/ds\), while the unipolar slope is \(-c\, dU/ds\) (Equation 6). Their ratio gives the conduction velocity without measuring activation times (Riccio et al. 2021; Deno et al. 2017):

\[ c \approx \frac{\mathrm{pp}\left( dU/dt \right)}{\mathrm{pp}\left( \hat n \cdot \hat{\mathbf E} \right)} \tag{16}\]

where pp is the peak-to-peak amplitude. Module m25 returns to this.

Where it breaks down. The estimate assumes the wave is planar and the potential is linear inside the clique. Neither holds when the spacing is not small next to \(a = w + h\). In our model, without time alignment, the spread of omnipolar amplitude across 18 wave directions grows from under 1% at \(d/a = 0.2\) to about 32% at \(d/a = 1.6\) for a square clique, and more for a triangle (computed with labs/python/epml; the trend is checked in tests/test_m03_omnipolar.py). Published work reports the same mechanism: bipoles within a clique activate at slightly different times, which biases the estimate, and aligning them first reduces the bias (Riccio et al. 2021). Curved or colliding wavefronts make the field trace a true loop, and no single direction then describes propagation.

Approximate model

Assumptions

  • Model: the planar arctangent front crossing a square clique of point electrodes. The field comes from all six bipoles by least squares (Equation 14), and the omnipole reads it along the loop’s principal axis (Equation 15).
  • The published method picks the direction by cross-correlation with the unipolar derivative and can time-align the bipoles first (Riccio et al. 2021). Neither is done here, which is why the omnipole is not perfectly direction-independent at larger spacing.
  • Default spacing 2 mm, the grid spacing in a published omnipolar simulation, and height 1 mm from the same study (Riccio et al. 2021); front width as in m03-d1. With these defaults \(d/a\) is about 1.5. Reduce the spacing to see the omnipole become direction-independent.
  • What would change the result: time alignment, curved wavefronts (the loop opens up) and noise.

Key takeaways

  • A clique of electrodes estimates the local electric field by least squares.
  • Reading that field along the propagation direction removes most orientation dependence.
  • The method assumes a locally planar wave; it degrades as spacing grows relative to \(w + h\).

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.

Why the same tissue can look normal or low voltage. Bipolar voltage maps are used to outline scar, for example in studies of ventricular tachycardia (a fast rhythm that starts in the lower chambers of the heart) (Marchlinski et al. 2000). In that study, normal endocardium (the inner surface of the heart wall) was defined as a bipolar amplitude above 1.5 mV, based on maps in sinus rhythm (the normal heartbeat) from 6 patients without structural heart disease, and “dense scar” as below 0.5 mV (Marchlinski et al. 2000). Those values were set with the catheters and mapping system of that study. The same two values appear in later studies (Takigawa et al. 2018; Magtibay et al. 2017).

This module shows why a cut-off cannot be separated from the instrument. A close bipole’s amplitude is \(\frac{9A}{8a^2}\, d\, \lvert\cos\theta\rvert\): it depends on the source strength \(A\), but also on \(w + h\), the spacing \(d\) and the angle \(\theta\) to the wavefront, and the electrode size filters it further (see the cheat sheet). Published studies show each effect:

  • Orientation. At the same sites, two crossing bipoles of equal spacing put many points in different voltage ranges, most often between 0.5 and 1.5 mV (Takigawa et al. 2018).
  • Electrode size and spacing. Two catheters gave similar normal atrial values (fifth percentiles 0.48 and 0.52 mV), yet in scarred atria the smaller, closer electrodes measured a smaller low-voltage area (Anter et al. 2015).
  • Recording method. A study of omnipolar voltage set its own threshold from ablation lesions (small areas of tissue destroyed on purpose, here by radiofrequency heating) in pig hearts. Maps using the traditional bipolar dense-scar cut-off of 0.5 mV outlined those lesions poorly (Magtibay et al. 2017).

An editorial asks the question directly: what does bipolar voltage amplitude really mean (Anter and Josephson 2016)? The engineering answer from this module: it describes the tissue as seen by a particular catheter, at a particular orientation, for a particular wavefront. Module m23 covers voltage mapping.

Visual summary

Schematic

1 · Tissue
Membrane current \(i_m\): a source ahead of the front (+), a sink behind it (−).

2 · Height
\(e^{-\lvert k\rvert h}\): low-pass. Lifting the electrode blurs fine detail.

3 · Electrode size
\(2J_1(kr_e)/(kr_e)\): low-pass. A larger electrode averages more.

4 · Bipole
\(2\lvert\sin(kd\cos\theta/2)\rvert\): high-pass, with nulls. Depends on the angle \(\theta\).

5 · Recorded bipolar signal
The product: a band-pass view of the tissue.

Figure 1: From tissue to recorded bipolar electrogram. Each stage multiplies the spatial spectrum of the membrane current by its transfer function; the small curves show each factor against spatial frequency \(k\).

Assumptions

  • Schematic: the curves use the module’s formulas with illustrative parameter values (Equation 4, Equation 11, Equation 12). They are not to scale and make no quantitative claim. The tissue sketch shows the shape of \(\partial^2 V_m/\partial x^2\) for the arctangent front (Equation 3).
  • The chain is the module’s model: a thin tissue sheet, an infinite homogeneous conductor and a planar front. Demo m03-d3 plots the same factors with real units.

How experts think

Treat every electrogram as a measurement of the tissue through an instrument. Before reading meaning into its amplitude or timing, ask:

  1. What is it referenced to? A unipolar signal is measured against a distant electrode, so it includes far-field activity. For a bipolar signal, which electrode is distal? The sign depends on it (Equation 8).
  2. Which way was the wave going? A bipole that lies across the wavefront records little, whatever the tissue (Equation 10). A second, crossing bipole or an omnipole (Equation 15) answers the question.
  3. What were the spacing, electrode size and height? Together they set which spatial scales survive (cheat sheet, spatial filters).
  4. Is the signal local? In this module’s model, a close bipole peaks when the unipolar signal at its center falls fastest (Q6 above). Unipolar recordings taken at the same time are a useful check on bipolar ones (Stevenson and Soejima 2005).
  5. What filtering was applied? High-pass filtering changes the shape of a unipolar signal (Stevenson and Soejima 2005). Module m04 covers this.
  6. Do the model’s assumptions hold here? The markers derived in this module assume planar, uniform propagation. Near changes in activation sequence, curved fronts or non-uniform coupling they can mislead (Steinhaus 1989; Riccio et al. 2021).

One sentence to keep: a bipolar voltage is a property of the tissue, the catheter and the wavefront together.

Cheat sheet

Everything here uses this module’s model: a thin tissue sheet in an infinite homogeneous conductor, a planar front with the arctangent profile, and point electrodes unless stated. Throughout, \(a = w + h\), and \(s\) is the distance from the front to the electrode, positive before the front arrives.

Signals

Quantity Result
Unipolar potential (Equation 1) \(\phi = \frac{1}{4\pi\sigma} \iint \frac{i_m}{\lVert \mathbf r - \mathbf r' \rVert}\, dA'\)
Membrane current (Equation 2) \(i_m = \beta\, \partial^2 V_m / \partial x^2\)
Unipolar, planar front (Equation 5) \(U(s) = A\, s/(s^2 + a^2)\). Peaks at \(s = \pm a\), zero at \(s = 0\), peak-to-peak \(A/a\)
Unipolar timing (Equation 6) Peak to trough takes \(2a/c\). Steepest slope \(-cA/a^2\), at \(s = 0\)
Activation marker (Equation 7) \(t_{\text{act}} \approx \arg\max_t(-dU/dt)\). Exact only for ideal propagation
Bipolar (Equation 8) \(V_{\text{bip}} = \phi(\mathbf r_1) - \phi(\mathbf r_2)\), distal minus proximal
Close bipole, \(d \ll a\) (Equation 10) \(V_{\text{bip}} \approx d\cos\theta\, \dfrac{dU}{ds} = -\dfrac{d \cos\theta}{c}\, \dfrac{dU}{dt}\)
Close-bipole amplitude Peak-to-peak \(\to \dfrac{9A}{8a^2}\, d\, \lvert\cos\theta\rvert\)
Clique field (Equation 14) \(\hat{\mathbf E} = -(D D^{\mathsf T})^{-1} D\, \mathbf b\). Needs three or more electrodes not in a line
Omnipole (Equation 15) \(v_\omega = -d\, \hat\omega \cdot \hat{\mathbf E}\)
Velocity from one clique (Equation 16) \(c \approx \mathrm{pp}(dU/dt) / \mathrm{pp}(\hat n \cdot \hat{\mathbf E})\)

Spatial filters (spatial frequency \(k\) in rad/mm)

Element Transfer function Behavior
Height \(h\) \(e^{-\lvert k\rvert h}\) (Equation 4) Low-pass
Disk electrode, radius \(r_e\) \(2 J_1(k r_e)/(k r_e)\) (Equation 11) Low-pass; first null at \(k r_e \approx 3.83\)
Bipole, spacing \(d\) \(2\lvert\sin(k d \cos\theta / 2)\rvert\) (Equation 12) High-pass; peak at \(k d\cos\theta = \pi\), nulls at multiples of \(2\pi\)
Bipolar recording Product of all three Band-pass

Far field from a source at distance \(R \gg d\) (Equation 13): for a monopole, the unipole falls as \(1/R\) and the bipole as \(d/R^2\); for a dipole, as \(1/R^2\) and \(d/R^3\).

What changes what (close bipole, \(d \ll a\))

Change Unipolar peak-to-peak Unipolar steepest slope Unipolar peak-to-trough time Close-bipole peak-to-peak
Double \(a = w + h\) Halves Quarters Doubles Quarters
Double \(c\) Same Doubles Halves Same
Double \(d\) Same Same Same Doubles
Turn the bipole to angle \(\theta\) Same Same Same Times \(\lvert\cos\theta\rvert\)

Rules of thumb

  • A unipole sees near and far. A bipole sees near, along its own axis.
  • In this model a small bipolar voltage can come from a weak source (small \(A\)), a large \(w + h\), a short spacing, or a wave crossing the bipole near 90°. The voltage alone cannot tell these apart.
  • Height and electrode size blur. Spacing decides which spatial scales a bipole keeps.
  • An omnipole removes orientation dependence only while the clique is small next to \(w + h\) and the wave is locally planar.
  • Units: 1 m/s = 1 mm/ms, so a length in mm divided by a speed in m/s gives a time in ms.

Interview questions

Try each question before you open its model answer. The answers use this module’s model and symbols. In a real interview, say which assumptions you are making.

Fundamentals

Q1. What does a unipolar electrogram measure, and why does a planar wavefront passing under the electrode give a positive deflection followed by a negative one?

Model answer

It measures the potential at one electrode relative to a distant reference. By Equation 1, that potential is a sum of membrane currents weighted by one over distance. Ahead of the front, membrane current flows outward (a source); behind it, current flows inward (a sink), and the two balance over the whole front. While the front approaches, the source side is closer, so the potential is positive. When the front is directly under the electrode, the two sides cancel and the potential crosses zero. After the front passes, the sink side is closer, so the potential is negative. In closed form, \(U(s) = A s/(s^2 + a^2)\): peaks at \(s = \pm a\), zero at \(s = 0\) (Equation 5).

Q2. Why is the steepest negative slope of the unipolar signal used as the local activation time, and when does that marker become unreliable?

Model answer

At a fixed electrode \(dU/dt = -c\, dU/ds\) (Equation 6). For the planar front \(dU/ds\) is largest at \(s = 0\), so the steepest downstroke happens exactly when the front passes under the electrode (Equation 7). That holds for ideal, uniform propagation. Simulation work found small errors under uniform conditions and larger ones near changes in activation sequence, in short strands and with non-uniform coupling (Steinhaus 1989). Curved or colliding fronts break the planar-wave assumption the derivation relies on.

Q3. The electrode is lifted so that \(w + h\) doubles. What happens to the unipolar amplitude, the amplitude of a close bipole, and the signal duration? Which recording is more local, and why?

Model answer

Unipolar peak-to-peak is \(A/a\), so it halves. A close bipole’s peak-to-peak is \(\frac{9A}{8a^2}\, d\, \lvert\cos\theta\rvert\), so it falls to a quarter. The unipolar peak-to-trough time is \(2a/c\), so it doubles: the signal gets smaller and broader. The bipole is more local: its amplitude falls off faster with distance, so tissue close to the catheter dominates it. In spatial-frequency terms, height multiplies the spectrum by \(e^{-\lvert k\rvert h}\), and the bipole mostly keeps the higher frequencies that this factor removes first (Equation 4).

Q4. Describe a bipolar recording as a spatial filter. What sets its low and high edges, and where are its nulls?

Model answer

A bipolar recording sees the tissue’s current pattern through the product \(H_{\text{bip}} \cdot H_{\text{disk}} \cdot e^{-\lvert k\rvert h}\). The height factor and the disk average \(2J_1(kr_e)/(kr_e)\) are low-pass filters, so height and electrode size set the high edge: fine detail is blurred, and the disk factor first reaches zero at \(k r_e \approx 3.83\). The bipole factor \(2\lvert\sin(kd\cos\theta/2)\rvert\) is zero at \(k = 0\), so spacing sets the low edge. It peaks at \(kd\cos\theta = \pi\) and has nulls wherever \(kd\cos\theta\) is a multiple of \(2\pi\) (Equation 12). Together they make a band-pass filter, and orientation enters through \(\cos\theta\).

Derivation questions

Q5. Show that a close bipole measures a directional derivative. How large is the error of that approximation, and what does it predict about catheter orientation?

Model answer

For a planar wave the two electrodes sit at \(s \pm \delta\), with \(\delta = \tfrac{d}{2}\cos\theta\) (Equation 9). A Taylor expansion gives \(U(s + \delta) - U(s - \delta) = 2\delta\, U'(s) + \tfrac{\delta^3}{3} U'''(s) + \dots\), so \(V_{\text{bip}} \approx d\cos\theta\, dU/ds = d\, \hat u \cdot \nabla\phi\) (Equation 10). The next term is smaller by a factor of order \(\delta^2\), so the relative error falls as \(d^2\). The amplitude scales with \(\lvert\cos\theta\rvert\): a bipole aligned with the wave records the most, and one at 90° records nothing, because both electrodes activate together. In animal recordings, voltage at the same site varied most where the wavefront was perpendicular to one bipole and parallel to the other (Takigawa et al. 2018).

Q6. Show that, in this model, a close bipolar electrogram is proportional to the time derivative of the unipolar signal at its center. When does its largest value occur?

Model answer

Combine \(V_{\text{bip}} \approx d\cos\theta\, dU/ds\) (Equation 10) with \(dU/dt = -c\, dU/ds\) (Equation 6):

\(V_{\text{bip}}(t) \approx -\dfrac{d\cos\theta}{c}\, \dfrac{dU}{dt}\),

where \(U\) is the unipolar signal at the bipole’s center. The bipolar signal is largest in magnitude when \(-dU/dt\) is largest, which is when the front passes under the center of the bipole. So in the ideal case, the bipolar peak and the unipolar steepest downstroke mark the same moment. The factor \(1/c\) cancels the factor \(c\) in the slope, so the bipolar amplitude does not depend on conduction velocity.

Q7. Why do bipoles reject far-field activity better than unipoles? Derive the decay exponents, and say why the rejection is never complete.

Model answer

A distant source changes slowly across the catheter, so a bipole records about \(d\) times the gradient of its potential. A monopole’s potential falls as \(1/R\), so its gradient, and the bipole signal, fall as \(1/R^2\). A dipole’s potential falls as \(1/R^2\), so the bipole signal falls as \(1/R^3\) (Equation 13). The bipole is always one power of \(R\) better than the unipole. But the gradient of a distant source is small, not zero, so far-field signals still appear in bipolar recordings, and the local signal can lie under either electrode (Stevenson and Soejima 2005).

Q8. How do you estimate the local electric field from a clique of electrodes, and what is the smallest clique that works?

Model answer

If the potential is nearly linear across the clique, each bipole is a projection of the field: \(b_{ij} = \phi_j - \phi_i \approx -(\mathbf r_j - \mathbf r_i) \cdot \mathbf E\). Stacking the bipoles gives \(\mathbf b = -D^{\mathsf T}\mathbf E\), and least squares gives \(\hat{\mathbf E} = -(DD^{\mathsf T})^{-1} D\, \mathbf b\) (Riccio et al. 2021) (Equation 14). A field in the plane has two unknowns, so the clique needs two independent bipole directions: at least three electrodes that are not in a line. If the electrodes are collinear, \(DD^{\mathsf T}\) has rank 1, and only the field component along the line can be measured. A square clique gives six bipoles for the same two unknowns, and least squares combines them.

Q9. Derive the conduction velocity from a single clique, without measuring activation times. Check the units.

Model answer

Along the propagation direction the field is \(\hat n \cdot \mathbf E = -dU/ds\), and the unipolar slope is \(dU/dt = -c\, dU/ds\) (Equation 6). Dividing one by the other cancels \(dU/ds\), so \(c \approx \mathrm{pp}(dU/dt)/\mathrm{pp}(\hat n \cdot \hat{\mathbf E})\) (Equation 16) (Riccio et al. 2021; Deno et al. 2017). Units: (mV/ms) divided by (mV/mm) is mm/ms, which equals m/s. The estimate assumes a plane wave across the clique and that the unipolar signal is recorded at the clique. Module m25 returns to conduction velocity.

Design and troubleshooting

Q10. At one site, the bipolar amplitude halves after the catheter is rotated, and nothing else changes. Give a model explanation, and say how you would tell an orientation effect from a real change in the tissue.

Model answer

For a close bipole the amplitude scales with \(\lvert\cos\theta\rvert\) (Equation 10). If the bipole was aligned with the wave before, halving means \(\lvert\cos\theta\rvert\) went from 1 to 0.5, which is a rotation of 60° relative to the propagation direction. To separate orientation from tissue: look at the unipolar amplitude, which a point electrode records independently of catheter orientation; compare several bipole directions at the same site; or use an omnipolar estimate, which, within its assumptions, reads the field along the propagation direction whatever the catheter’s orientation (Equation 15).

Q11. For a fixed local wavefront and a fixed distant source, how does bipole spacing change the share of far-field signal in a bipolar recording?

Model answer

While \(d \ll a\), both the local and the far-field bipolar signals are about \(d\) times a gradient, so their ratio barely depends on \(d\). Once \(d\) is comparable to \(a = w + h\), the local signal stops growing: the bipole spans the whole front, and for very wide spacing it records two separate unipolar deflections, with peak-to-peak tending to \(A/a\). The far-field term keeps growing with \(d\) as long as \(d \ll R\). So wider spacing raises the far-field share. In this model, with a monopole at \(R = 200a\) on the bipole axis, the share stays within 1% of the short-bipole value up to \(d = 0.2a\). It is about 1.2 times that value at \(d = a\), and about 1.9 times at \(d = 2a\).

Q12. Why is an omnipolar amplitude independent of catheter orientation, and which assumptions break that?

Model answer

The clique estimates the field vector \(\hat{\mathbf E}\), whichever bipoles the catheter happens to offer. Reading it along the propagation direction gives the signal of a bipole aligned with the wave, whatever the catheter’s orientation (Equation 15). That relies on a locally planar wave and a potential that is linear across the clique. Both fail as the spacing grows relative to \(w + h\). In this model, without time alignment, the spread of omnipolar amplitude across wave directions grows from under 1% at \(d/a = 0.2\) to about 32% at \(d/a = 1.6\) for a square clique. Published work traces the same bias to bipoles in a clique activating at slightly different times, and aligning them first reduces it (Riccio et al. 2021). Curved or colliding fronts make the field trace a true loop, and then no single direction describes propagation.

Q13. In this model, does slower conduction lower unipolar or bipolar voltage? What does it change?

Model answer

Not on its own. With \(w\), \(h\) and the source strength fixed, the signal is \(U(x_e - ct)\), so \(c\) only rescales the time axis. Unipolar peak-to-peak stays \(A/a\), and close-bipole peak-to-peak stays \(\frac{9A}{8a^2}\, d\, \lvert\cos\theta\rvert\). What changes is timing: the unipolar peak-to-trough time \(2a/c\) grows, and the steepest slope \(cA/a^2\) falls, as \(c\) falls. In this model, a voltage change has to come from \(A\), \(w\), \(h\), \(d\) or \(\theta\). The model treats those as independent of \(c\). Whether they stay fixed when conduction slows in real tissue depends on why it slows; module m09 covers propagation.

Code lab

The lab rebuilds this module’s results in code, so you can see where each one comes from and break it on purpose. It runs in about a minute.

Part What you do Data
A Compute a unipolar electrogram by summing the volume-conductor integral, check it against Equation 5, and find the activation marker synthetic
B Rotate a bipole and measure when the \(\lvert\cos\theta\rvert\) rule holds synthetic
C Treat height and the bipole as spatial filters with FFTs (Equation 4, Equation 12) synthetic
D Estimate the field from a clique, build an omnipole, and get conduction velocity from one clique (Equation 16) synthetic
E Look at real bipolar recordings from a decapolar catheter and estimate delays between neighboring bipoles PhysioNet

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. A planar front moves at 1.0 m/s past an electrode with \(w + h\) = 1.3 mm (the module’s defaults). How long after the positive peak of the unipolar electrogram does its negative trough come?

Answer

\(2a/c\) = 2 × 1.3 mm ÷ 1.0 mm/ms = 2.6 ms (1 m/s = 1 mm/ms). See the cheat sheet, unipolar timing.

S2. A close bipole aligned with the wave is turned by 60°. What fraction of its amplitude is left? What about at 90°?

Answer

\(\lvert\cos 60°\rvert\) = 0.5, so half. At 90° both electrodes activate together, and in this model the bipole records nothing (Equation 10).

S3. A 2-mm bipole lies along the propagation direction. Which spatial wavelength does it cancel completely, and which does it pass best?

Answer

\(\lvert H_{\text{bip}}\rvert = 2\lvert\sin(kd/2)\rvert\) is zero when \(kd = 2\pi\), which is a wavelength \(\lambda = 2\pi/k\) equal to the spacing: 2 mm (and 1 mm, 0.67 mm, …). It is largest when \(kd = \pi\), at \(\lambda = 2d\) = 4 mm (Equation 12).

S4. A distant point source moves twice as far away. By what factor do its unipolar and bipolar signals fall?

Answer

The unipolar signal falls as \(1/R\), so by 2. The bipolar signal falls as \(1/R^2\), so by 4 (Equation 13).

S5. A disk electrode has radius 0.5 mm (demo m03-d3’s default). At what spatial wavelength does its averaging first cancel a pattern completely?

Answer

The disk factor \(2J_1(kr_e)/(kr_e)\) first reaches zero at \(kr_e \approx 3.83\), so \(k \approx 7.66\) rad/mm, and \(\lambda = 2\pi/k \approx 0.82\) mm (Equation 11).

S6. Which of these does the single-clique estimate of conduction velocity not need? (a) the unipolar slope \(dU/dt\); (b) the field along the propagation direction; (c) local activation times; (d) a wave that is planar across the clique.

Answer

(c). The estimate is the ratio \(\mathrm{pp}(dU/dt)/\mathrm{pp}(\hat n \cdot \hat{\mathbf E})\), which needs no activation times, but it does assume a planar wave across the clique (Equation 16).

S7. True or false: in this model, halving the conduction velocity halves the bipolar voltage.

Answer

False. With \(w\), \(h\) and the source fixed, \(c\) only stretches time: the voltage stays the same, the signal lasts twice as long, and slopes halve (interview Q13).

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.

  • Electric-field loops with and without time alignment. Riccio and colleagues, Figure 5 [Riccio et al. (2021); free full text]. It shows bipolar electrograms and field loops for square and triangular cliques, with and without time alignment of the bipoles, for plane and curved wavefronts at three catheter orientations. Compare plane with curved fronts, and aligned with unaligned bipoles. In this module a plane wave traces a line along the propagation direction (Equation 15), and spacing that is large next to \(w + h\) distorts it (demo m03-d5).
  • Bipolar voltage maps in two orientations. Magtibay and colleagues, Figure 2 [Magtibay et al. (2017); free full text]. Bipolar electrograms and voltage maps from a rabbit, a pig and a human heart, along two perpendicular electrode orientations, with maps built from the larger bipole. Compare the two orientations for the same pacing site (the place the heart was stimulated from). This module’s explanation is the \(\lvert\cos\theta\rvert\) rule (Equation 10, demo m03-d2, interview Q10).
  • Thresholds and lesions. Magtibay and colleagues, Figure 8 (Magtibay et al. 2017). Omnipolar voltage maps with a threshold set from ablation lesions in pig hearts, compared with maps using the traditional bipolar dense-scar cut-off of 0.5 mV. It goes with the clinical correlate above.
  • The upstroke behind the front model. Loose and colleagues, Figure 3 [Loose et al. (2014); free full text, CC BY 3.0]. Original action-potential recordings from human right atrial tissue, one preparation per patient group. Look at how brief the upstroke is next to the rest of the action potential. Equation 3 keeps only the upstroke, compressed into a front of width \(w\); the m03-d1 assumptions derive \(w\) from this study’s amplitude.

Further reading

  • Theory. Plonsey and Barr (2007): a textbook on bioelectricity, including the volume-conductor theory behind Equation 1.
  • Recording practice. Stevenson and Soejima (2005): a short review of unipolar and bipolar recording, filtering and sampling in the electrophysiology lab. Bakker (2019): a review of electrogram recording and analysis techniques for choosing ablation target sites.
  • Omnipolar signals. Deno et al. (2017): the engineering paper that introduced omnipolar signals from electric-field loops and tested them in a saline tank. Massé et al. (2016): omnipolar conduction velocity checked against optical and electrical mapping in several models. Riccio et al. (2021): a simulation study of omnipolar strategies with square and triangular cliques, time alignment and noise. Magtibay et al. (2019): a review of what omnipolar electrograms add to mapping.
  • Voltage. Anter and Josephson (2016): an editorial on what bipolar voltage amplitude means.

References

Anter, Elad, and Mark E. Josephson. 2016. “Bipolar Voltage Amplitude: What Does It Really Mean?” Heart Rhythm 13 (1): 326–27. https://doi.org/10.1016/j.hrthm.2015.09.033.
Anter, Elad, Cory M. Tschabrunn, and Mark E. Josephson. 2015. “High-Resolution Mapping of Scar-Related Atrial Arrhythmias Using Smaller Electrodes with Closer Interelectrode Spacing.” Circulation: Arrhythmia and Electrophysiology 8 (3): 537–45. https://doi.org/10.1161/CIRCEP.114.002737.
Bakker, Jacques M. T. de. 2019. “Electrogram Recording and Analyzing Techniques to Optimize Selection of Target Sites for Ablation of Cardiac Arrhythmias.” Pacing and Clinical Electrophysiology 42 (12): 1503–16. https://doi.org/10.1111/pace.13817.
Deno, D. Curtis, Ram Balachandran, Dennis Morgan, Faisal Ahmad, et al. 2017. “Orientation-Independent Catheter-Based Characterization of Myocardial Activation.” IEEE Transactions on Biomedical Engineering 64 (5): 1067–77. https://doi.org/10.1109/TBME.2016.2589158.
Goldberger, A. L., L. A. Amaral, L. Glass, et al. 2000. “PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals.” Circulation 101 (23): e215–20. https://doi.org/10.1161/01.cir.101.23.e215.
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.
Loose, Simone, Judith Mueller, Erich Wettwer, et al. 2014. “Effects of IKur Blocker MK-0448 on Human Right Atrial Action Potentials from Patients in Sinus Rhythm and in Permanent Atrial Fibrillation.” Frontiers in Pharmacology 5: 26. https://doi.org/10.3389/fphar.2014.00026.
Magtibay, Karl, Stéphane Massé, John Asta, Marjan Kusha, et al. 2017. “Physiological Assessment of Ventricular Myocardial Voltage Using Omnipolar Electrograms.” Journal of the American Heart Association 6 (8). https://doi.org/10.1161/JAHA.117.006447.
Magtibay, Karl, Andreu Porta-Sánchez, Shouvik K. Haldar, et al. 2019. “Reinserting Physiology into Cardiac Mapping Using Omnipolar Electrograms.” Cardiac Electrophysiology Clinics 11 (3): 525–36. https://doi.org/10.1016/j.ccep.2019.05.003.
Marchlinski, F. E., D. J. Callans, C. D. Gottlieb, and E. Zado. 2000. “Linear Ablation Lesions for Control of Unmappable Ventricular Tachycardia in Patients with Ischemic and Nonischemic Cardiomyopathy.” Circulation 101 (11): 1288–96. https://doi.org/10.1161/01.cir.101.11.1288.
Massé, Stéphane, Karl Magtibay, Nicholas Jackson, John Asta, et al. 2016. “Resolving Myocardial Activation with Novel Omnipolar Electrograms.” Circulation: Arrhythmia and Electrophysiology 9 (7): e004107. https://doi.org/10.1161/CIRCEP.116.004107.
Plonsey, Robert, and Roger C. Barr. 2007. Bioelectricity: A Quantitative Approach. 3rd ed. Springer. https://doi.org/10.1007/978-0-387-48865-3.
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.
Stevenson, William G., and Kyoko Soejima. 2005. “Recording Techniques for Clinical Electrophysiology.” Journal of Cardiovascular Electrophysiology 16 (9): 1017–22. https://doi.org/10.1111/j.1540-8167.2005.50155.x.
Takigawa, Masateru et al. 2018. “Effect of Bipolar Electrode Orientation on Local Electrogram Properties.” Heart Rhythm, ahead of print. https://doi.org/10.1016/j.hrthm.2018.07.020.
Wang, G. Y., J. Wei, and C. I. Lin. 1991. “Depressant Effect of Adenosine in Isolated Human and Canine Atrial Fibres.” Cardiovascular Research 25 (1): 31–35. https://doi.org/10.1093/cvr/25.1.31.