m23 · Voltage mapping: amplitude, thresholds, dependence on wavefront direction
Education only. Not for diagnosis, treatment or any clinical decision.
In 60 seconds
A voltage map colors each site by the peak-to-peak amplitude of its electrogram. Low amplitude is read as diseased tissue or scar, but the amplitude also depends on the recording. In the m03 model, the unipolar amplitude is \(A/a\), so it falls as the electrode lifts off. A closely spaced bipole records \(\tfrac98 A d\lvert\cos\theta\rvert/a^2\), so it falls faster with height and also depends on the angle between the bipole and the wavefront. If wavefront directions are random, healthy tissue reads below a third of its aligned amplitude about one time in five. The larger of two orthogonal bipoles never drops below \(1/\sqrt2\) of it, and the largest projection of the pair is nearly independent of direction at small spacing. Noise sets a floor: with no signal, the peak-to-peak of 100 noise samples is about five noise SDs. Cutoffs such as 0.5 mV and 1.5 mV then trade sensitivity for specificity, and contact moves healthy tissue toward them.
From the field
- The classic ventricular cutoffs. In an early study of 16 patients, normal ventricular endocardium was defined as bipolar amplitude above 1.5 mV, from sinus-rhythm maps of 6 patients without structural heart disease, and dense scar as below 0.5 mV (Marchlinski et al. 2000).
- Atrial normal values depend little on electrode size, scar maps more. In 10 patients with normal atria, the fifth percentile of bipolar voltage was 0.48 mV with 3.5-mm electrodes and 0.52 mV with 1-mm electrodes. In 20 patients with scar, the area below 0.5 mV was 20.4 cm² with the larger electrodes and 14.7 cm² with the smaller ones (Anter et al. 2015).
- Contact force and direction matter. In 14 patients with repaired tetralogy of Fallot, contact force correlated with unipolar and bipolar voltage (Teijeira-Fernandez et al. 2018). In 10 swine, bipolar voltage from a stationary array differed by 32% (13–53%) between orthogonal activation directions, but by 11% (5–18%) when the catheter roved (Yavin et al. 2022).
Derivations
Symbols used throughout. \(U\) unipolar potential with strength \(A\); \(s\) front distance; \(a = w + h\) front width plus electrode height (m03); \(\operatorname{pp}\) peak-to-peak; \(d\) bipole spacing; \(\theta\) angle between bipole and propagation; \(q\) a fraction of the aligned amplitude; \(n\) number of independent noise samples; \(\sigma\) noise SD; \(\phi\), \(\Phi\) standard normal density and distribution; \(t\) a voltage cutoff; \(m_s\), \(m_n\) median voltages of scar and healthy tissue; \(\sigma_s\), \(\sigma_n\) SDs of their logarithms.
Amplitude
From the m03 front, \(U = As/(s^2 + a^2)\) peaks at \(s = \pm a\), and \(dU/ds\) ranges from \(A/a^2\) to \(-A/(8a^2)\):
\[ \operatorname{pp}(U) = \frac{A}{a}, \qquad \operatorname{pp}(B) \to \frac98\,\frac{A\,d\lvert\cos\theta\rvert}{a^2} \;\; (d\lvert\cos\theta\rvert \ll a), \qquad \operatorname{pp}(B) \to \frac{A}{a} \;\; (d\lvert\cos\theta\rvert \gg a) \tag{1}\]
A close bipole is a directional derivative (m03, eq-m03-bipolar-gradient), so its amplitude falls as \(1/a^2\), faster than the unipolar \(1/a\). Lifting the electrode from 1 to 2 mm kept 0.57 of the unipolar amplitude, 0.42 of a 2-mm bipole’s and 0.32 of a 0.2-mm bipole’s (lab part A). For a 2-mm bipole at 1 mm height, the small-spacing formula overstates the amplitude by half, so use the exact form there.
Key takeaways
- Bipolar voltage reflects contact more strongly than unipolar.
- Cutoffs belong to a catheter: spacing and electrode size change the reading.
Direction
In the small-spacing limit a bipole reads \(\lvert\cos\theta\rvert\) times its aligned amplitude. For uniformly random in-plane directions:
\[ P\bigl(\lvert\cos\theta\rvert < q\bigr) = \frac{2}{\pi}\arcsin q, \qquad \mathrm E\lvert\cos\theta\rvert = \frac2\pi, \qquad \max\bigl(\lvert\cos\theta\rvert, \lvert\sin\theta\rvert\bigr) \ge \frac{1}{\sqrt2} \tag{2}\]
With \(q = 1/3\), the ratio of the 0.5-mV and 1.5-mV cutoffs, healthy tissue exactly at the normal cutoff reads as dense scar 21.6% of the time. In the lab, a 2-mm bipole did so 14% of the time; finite spacing softens the cosine. The larger of two orthogonal bipoles never fell below the cutoff. The largest projection of the pair, the idea behind omnipolar methods (m03) (Massé et al. 2016; Deno et al. 2017), was independent of direction at 0.2-mm spacing and varied by up to 20% at 2 mm (lab part B). A simulation found modified omnipolar maps more accurate than bipolar maps at detecting fibrosis (Riccio et al. 2021).
Exact
Assumptions
- Model: the m03 arctangent front, planar, crossing healthy tissue; peak-to-peak amplitudes computed exactly on the closed-form signals of one bipole and of two orthogonal bipoles sharing a center (Equation 1, Equation 2). Tested against
epml.voltage. - Parameters: front width 0.3 mm (m03); default height 1 mm (Riccio et al. 2021) and spacing 2 mm (Intracardiac Atrial Fibrillation Database (Version 1.0.0) 2003), as in m03; the default threshold 0.33 is the ratio of the 0.5-mV and 1.5-mV cutoffs (Marchlinski et al. 2000).
- What it ignores: directions out of the tissue plane, curved fronts, noise (Equation 3), heterogeneous tissue and electrode size.
Noise floor
With no local signal, the peak-to-peak of \(n\) independent noise samples is still positive:
\[ \mathrm E[\operatorname{pp}] = 2\sigma\int_{-\infty}^{\infty} x\, n\,\phi(x)\,\Phi(x)^{n-1}\,dx \tag{3}\]
That is about \(3.1\sigma\) for 10 samples, \(5.0\sigma\) for 100 and \(6.5\sigma\) for 1000 (lab part C). With an illustrative noise SD of 0.02 mV and 100 independent samples in the window, the floor is 0.10 mV: nothing below it can be mapped. Filtering makes samples dependent, so \(n\) is the number of independent samples, not of raw samples.
Cutoffs
Model the voltages of scar and healthy tissue as log-normal. A cutoff \(t\) then gives:
\[ \text{sensitivity} = \Phi\Bigl(\frac{\ln t - \ln m_s}{\sigma_s}\Bigr), \qquad \text{specificity} = 1 - \Phi\Bigl(\frac{\ln t - \ln m_n}{\sigma_n}\Bigr), \qquad t^* = \sqrt{m_s m_n} \text{ if } \sigma_s = \sigma_n \tag{4}\]
The last result maximizes sensitivity plus specificity. With illustrative medians of 0.2 and 2.5 mV and a log SD of 0.7, a 1.5-mV cutoff caught 99.8% of scar but called 23% of healthy tissue low. A 1-mm lift of a 2-mm bipole scaled every reading by 0.42 and raised that to 70% (lab part D). Cutoffs found by ROC analysis against imaging depend on the catheter, the chamber and contact: one study found 1.76 mV bipolar and 5.19 mV unipolar after excluding very low contact force, with sensitivity and specificity of 0.90/0.90 and 0.89/0.85 (Teijeira-Fernandez et al. 2018).
Approximate model
Assumptions
- Model: log-normal voltages for scar and healthy tissue with equal spread (Equation 4); extra height scales every reading by the exact factor for an aligned 2-mm bipole (Equation 1). Tested against
epml.voltage. - Parameters: medians 0.2 and 2.5 mV and the spread are illustrative; the default cutoff 1.5 mV is the classic normal cutoff (Marchlinski et al. 2000); base height 1 mm (Riccio et al. 2021).
- What it ignores: direction (Equation 2), noise (Equation 3), border-zone tissue between the two classes, and correlation between neighboring points.
Key takeaways
- A cutoff is a choice of operating point, not a property of tissue.
- Poor contact shifts healthy tissue toward low voltage; direction does too.
Clinical correlate
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. Voltage maps are used to outline scar and diseased tissue. The classic ventricular cutoffs came from a small series (Marchlinski et al. 2000). Atrial normal values have been measured for different electrode sizes (Anter et al. 2015). Contact force (Teijeira-Fernandez et al. 2018) and activation direction, especially for a stationary catheter (Yavin et al. 2022), change the voltage recorded from the same tissue.
What the evidence is. Clinical series (Marchlinski et al. 2000; Anter et al. 2015), a study against MRI (Teijeira-Fernandez et al. 2018), an animal study (Yavin et al. 2022) and a simulation study (Riccio et al. 2021).
What a reader should ask:
- Which catheter, spacing and electrode size? (Equation 1)
- Was contact controlled? (Equation 1)
- Were points collected from many directions or one? (Equation 2)
- What is the noise floor of the recording? (Equation 3)
- How were the cutoffs chosen, and against what reference? (Equation 4)
Visual summary
Schematic
Assumptions
- Schematic: not to scale, no quantitative claim. Labels follow Equation 1, Equation 2 and Equation 4.
How experts think
Before reading low voltage as scar, ask:
- Was the electrode in contact? (Equation 1)
- Could the wavefront have run across the bipole? (Equation 2)
- Is the amplitude above the noise floor? (Equation 3)
- Which cutoff, for which catheter, validated against what? (Equation 4)
- Does the low-voltage region persist with another rhythm or direction? (Equation 2)
One sentence to keep: a voltage map measures tissue, contact, direction and noise together.
Cheat sheet
| Quantity | Result |
|---|---|
| Amplitude (Equation 1) | \(\operatorname{pp}(U) = A/a\); \(\operatorname{pp}(B) \to \tfrac98 A d\lvert\cos\theta\rvert/a^2\) (small \(d\)), \(\to A/a\) (large \(d\)) |
| Direction (Equation 2) | \(P(\lvert\cos\theta\rvert < q) = \tfrac2\pi\arcsin q\); pair \(\ge 1/\sqrt2\) |
| Noise (Equation 3) | \(\mathrm E[\operatorname{pp}] = 2\sigma\,\mathrm E[\max_n]\): about \(5\sigma\) for 100 samples |
| Cutoffs (Equation 4) | log-normal sensitivity and specificity; \(t^* = \sqrt{m_s m_n}\) for equal spreads |
Values used in this module
| Value | Source |
|---|---|
| Normal > 1.5 mV, dense scar < 0.5 mV (ventricle) | Marchlinski et al. (2000) |
| Atrial fifth percentile 0.48 mV (3.5 mm) and 0.52 mV (1 mm); low-voltage area 20.4 vs 14.7 cm² | Anter et al. (2015) |
| Cutoffs 1.76 mV bipolar, 5.19 mV unipolar against MRI; 14 patients | Teijeira-Fernandez et al. (2018) |
| Orthogonal-direction difference 32% stationary, 11% roving; 10 swine | Yavin et al. (2022) |
| Height 1 mm; spacing 2 mm | Riccio et al. (2021), Intracardiac Atrial Fibrillation Database (Version 1.0.0) (2003) |
| Tissue medians, spreads, noise SD, sample counts | illustrative |
| Amplitude ratios, direction fractions, noise floors, rates | 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. What does a voltage map show?
Model answer
The peak-to-peak amplitude of the electrogram at each site, unipolar or bipolar (Equation 1).
Q2. Why does bipolar voltage depend more on contact than unipolar?
Model answer
A close bipole measures a spatial derivative, which falls as \(1/a^2\), against \(1/a\) for unipolar (Equation 1).
Q3. Why can’t you transfer a voltage cutoff between catheters?
Model answer
Bipolar amplitude depends on spacing and electrode size (Equation 1); normal values have been measured per catheter (Anter et al. 2015).
Q4. A healthy site reads low on one bipole. What else could explain it?
Model answer
The wavefront ran nearly perpendicular to the bipole, the electrode was not in contact, or the reading is at the noise floor (Equation 2, Equation 1, Equation 3).
Q5. Derive the fraction of random directions in which a bipole reads below a fraction q of its aligned value.
Model answer
\(\lvert\cos\theta\rvert < q\) on an interval of length \(\pi - 2\arccos q\) out of \(\pi\), so \(\tfrac2\pi\arcsin q\) (Equation 2).
Q6. How do two orthogonal bipoles help?
Model answer
Their larger amplitude is at least \(1/\sqrt2\) of the aligned value; projecting the pair onto the best direction recovers it at small spacing (Equation 2) (Massé et al. 2016).
Q7. Does direction matter less for a roving catheter?
Model answer
In one animal study it did: 11% against 32% for a stationary array, because each site is sampled at several angles (Yavin et al. 2022).
Q8. What sets the lowest voltage a map can show?
Model answer
The peak-to-peak of noise in the measurement window, about five noise SDs for 100 independent samples (Equation 3).
Q9. How would you choose a cutoff?
Model answer
Against a reference such as imaging, with an ROC curve, choosing the operating point for the purpose (Equation 4) (Teijeira-Fernandez et al. 2018).
Q10. Two log-normal populations with equal spread. Where does sensitivity plus specificity peak?
Model answer
At the geometric mean of the medians (Equation 4).
Q11. Where did the 0.5-mV and 1.5-mV ventricular cutoffs come from?
Model answer
1.5 mV from sinus-rhythm maps of 6 patients without structural heart disease; 0.5 mV was used for dense scar (Marchlinski et al. 2000).
Q12. Smaller electrodes found less low-voltage area in atrial scar. Why might that be?
Model answer
They average over less tissue and record more distinct local signals, so surviving muscle reads higher (Anter et al. 2015); the m03 disk average shows the spatial filtering.
Code lab
The lab measures voltage on synthetic signals and a synthetic map.
| Part | What you do | Data |
|---|---|---|
| A | Amplitude, height and spacing (Equation 1) | synthetic |
| B | Wavefront direction (Equation 2) | synthetic |
| C | The noise floor (Equation 3) | synthetic |
| D | Cutoffs (Equation 4) | synthetic |
| E | A synthetic voltage map | synthetic |
- Python notebook: labs/python/m23_lab.ipynb. Run it from the course repository after
pip install -r requirements.txt, or open it in Google Colab. - Data: none to download.
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. \(a\) goes from 1 mm to 2 mm. Unipolar amplitude changes by what factor?
Answer
1/2 (Equation 1).
S2. Same change, a very close bipole?
Answer
1/4 (Equation 1).
S3. A close bipole at 60° to propagation reads what fraction of its aligned amplitude?
Answer
\(\cos 60° =\) 0.5 (Equation 2).
S4. Random directions; fraction reading below half of aligned?
Answer
\(\tfrac2\pi\arcsin 0.5 = 1/3\) (Equation 2).
S5. Two independent unit-normal samples. Expected peak-to-peak?
Answer
\(2/\sqrt\pi \approx\) 1.13 (Equation 3).
S6. Medians 0.25 and 4 mV, equal spread. Best cutoff for sensitivity plus specificity?
Answer
\(\sqrt{0.25 \times 4} =\) 1 mV (Equation 4).
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.
- Voltage and electrode size. Anter and colleagues’ paired maps with large and small electrodes (Anter et al. 2015). This module: amplitude.
- Direction in stationary and roving maps. Yavin and colleagues’ maps of one gap from four directions (Yavin et al. 2022). This module: direction.
- Voltage against MRI scar. Teijeira-Fernandez and colleagues’ ROC curves with and without contact-force filtering (Teijeira-Fernandez et al. 2018). This module: cutoffs.
Further reading
- Cutoffs. The classic ventricular values (Marchlinski et al. 2000); atrial values by electrode size (Anter et al. 2015); validation against MRI (Teijeira-Fernandez et al. 2018).
- Direction. Animal data (Yavin et al. 2022); omnipolar methods (Massé et al. 2016; Deno et al. 2017; Riccio et al. 2021).
- Related modules. m03 (electrograms), m13 (catheter design), m22 (annotation), m24 (interpolation), m26 (map quality), m40 (AI on imaging substrate), m52 (cardiac MRI).