Skip to main content

m19 · Motion and drift: respiration, cardiac gating, reference catheters

Warning

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

In 60 seconds

A mapping system wants the position of a point on a still heart, but the heart moves with every breath and every beat, and impedance coordinates slowly drift. Breathing is periodic and asymmetric, with more time near exhale; model it as \(-b\cos^{2n}(\pi t/T)\). Gating accepts points only near one phase: a window \(\delta\) around exhale keeps residual motion below \(\delta\) and accepts the fraction \(1 - (2/\pi)\arccos((\delta/b)^{1/(2n)})\) of the time. A reference catheter or other surrogate can be subtracted, but subtraction only helps when the correlation \(\rho\) exceeds \(\sigma_r/(2\sigma_x)\); scaling the surrogate by a fitted slope leaves \(\sigma_x\sqrt{1 - \rho^2}\) and never hurts. Drift is estimated from repeated visits to a landmark, with an error that shrinks as the visits spread in time. Averaging removes motion at the cost of delay: a window of one breath removes breathing and lags by half a breath.

From the field

  • Breathing changes the map. In 22 patients, left atrial maps collected regardless of respiratory phase overestimated pulmonary vein ostial dimensions compared with maps collected at end expiration, both judged against pre-procedural CT or MRI. Gated mapping took 7.8 ± 2.1 min against 3.2 ± 1.0 min (Beinart et al. 2011).
  • Surrogates need patient-specific scaling. In 12 volunteers imaged by real-time MR, the slope relating left main coronary to diaphragm displacement averaged about 0.6 but ranged from 0.17 to 0.93, with \(r^2\) from 0.04 to 0.87 (Danias et al. 1999).
  • Breathing is not a sine wave. A model used in radiotherapy treats breathing motion as periodic but asymmetric, with more time spent at exhale than at inhale (Lujan et al. 1999).
  • Drift over a procedure. With impedance-based localization, a landmark at the proximal His bundle was reproducible within 1.4 ± 1.1 mm after 2 hours (Wittkampf et al. 1999).

Derivations

Symbols used throughout. \(z(t)\) displacement from the exhale position (mm); \(b\) breathing excursion; \(T\) breathing period (s); \(n\) shape parameter; \(\delta\) gating window; \(D\) duty cycle; \(x\) a mapping point’s motion and \(\sigma_x\) its SD; \(r\) a reference’s motion and \(\sigma_r\) its SD; \(\rho\) their correlation; \(k\) regression slope; \(t_i\), \(x_i\) times and positions of landmark visits; \(v\) drift rate; \(\sigma\) measurement noise; \(W\) averaging window; \(f\) frequency.

Breathing

Breathing is periodic but asymmetric, with more time near exhale than inhale (Lujan et al. 1999). One simple model of the heart’s displacement from its exhale position:

\[ z(t) = -b\cos^{2n}\!\left(\frac{\pi t}{T}\right) \tag{1}\]

It is zero at exhale and \(-b\) at inhale. For \(n = 1\) it is a sinusoid with period \(T\); larger \(n\) flattens the exhale plateau. With an illustrative 12-mm excursion, the time within 1.2 mm of exhale rose from 20% at \(n = 1\) to 48% at \(n = 3\) (lab part A).

Gating

Accept a point only while \(\lvert z\rvert \le \delta\). Solving \(\cos^{2n}(\pi t/T) \le \delta/b\) over one period gives the accepted fraction of time:

\[ D = 1 - \frac{2}{\pi}\arccos\!\left[\left(\frac{\delta}{b}\right)^{1/(2n)}\right], \qquad \lvert z\rvert \le \delta \text{ for every accepted point} \tag{2}\]

Residual motion is bounded by the window, and the duty cycle is the price. With \(b\) = 12 mm and \(n\) = 2, a 2-mm window accepts 44% of the time with an RMS residual of 0.71 mm (lab part B). The same logic applies to cardiac gating on the ECG, with the cardiac cycle in place of the breath.

Exact

Assumptions

  • Model: Equation 1 with an illustrative 4-s period and 12-mm excursion; gating by Equation 2. Tested against epml.motion.
  • Parameters: the window and shape parameter are adjustable.
  • What it ignores: irregular breathing, cardiac motion, motion in three dimensions and the delay of the gating signal.

Key takeaways

  • Gating bounds residual motion by the window; the duty cycle pays for it.

Reference catheters and surrogates

Subtract a reference’s motion \(r\) from a mapping point’s motion \(x\):

\[ \operatorname{SD}(x - r) = \sqrt{\sigma_x^2 + \sigma_r^2 - 2\rho\sigma_x\sigma_r} < \sigma_x \iff \rho > \frac{\sigma_r}{2\sigma_x} \tag{3}\]

A reference that moves as much as the point helps only if their correlation exceeds 0.5. Scaling the surrogate first by the least-squares slope \(k = \rho\sigma_x/\sigma_r\) leaves:

\[ \operatorname{SD}(x - kr - c) = \sigma_x\sqrt{1 - \rho^2} \tag{4}\]

This never exceeds \(\sigma_x\). In the lab, with a true slope of 0.2, subtracting a diaphragm-like surrogate outright nearly tripled the motion (2.8 to 8.2 mm SD), while regression cut it to 2.0 mm (lab part C). Slopes vary widely between people (Danias et al. 1999), so they must be fitted per patient.

Exact

Assumptions

  • Model: jointly Gaussian motions with SDs and correlation as set; residual SDs by Equation 3 and Equation 4. Tested against epml.motion.
  • Parameters: correlation and the ratio of reference to point motion are adjustable.
  • What it ignores: time lags between the reference and the point, nonlinear relations and the reference catheter’s own displacement.

Key takeaways

  • Subtracting a reference can make things worse; regressing on it cannot.

Drift

Impedance coordinates can drift slowly. Revisit a landmark at times \(t_i\) and fit a line; the slope’s standard error is:

\[ \hat v = \frac{\sum_i (t_i - \bar t)(x_i - \bar x)}{\sum_i (t_i - \bar t)^2}, \qquad \operatorname{SE}(\hat v) = \frac{\sigma}{\sqrt{\sum_i (t_i - \bar t)^2}} \tag{5}\]

Spreading the same number of visits over a longer time shrinks the error: five visits over 120 minutes estimate drift four times better than five over 30 minutes (lab part D).

Averaging

Averaging position over a window \(W\) attenuates motion at frequency \(f\) and delays the result:

\[ \lvert H(f)\rvert = \left\lvert\frac{\sin(\pi f W)}{\pi f W}\right\rvert, \qquad \text{delay} = \frac{W}{2} \tag{6}\]

A window of exactly one breath removes breathing completely and lags by half a breath; half a breath leaves 64% of it (lab part E). Display smoothing is a trade between steadiness and lag.

Key takeaways

  • Drift is measured by revisiting landmarks, spread out in time.
  • Averaging removes motion at the cost of delay.

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. Respiratory compensation changes left atrial maps: in one study, maps collected at end expiration matched CT or MRI pulmonary vein ostia better than maps collected regardless of respiration, at the cost of a longer collection time (Beinart et al. 2011). The relation between the heart’s motion and surrogates such as the diaphragm varies between people (Danias et al. 1999). Landmark reproducibility over hours is one measure of drift (Wittkampf et al. 1999).

What the evidence is. A clinical comparison in 22 patients (Beinart et al. 2011), an MR study in 12 volunteers (Danias et al. 1999), a breathing-motion model from radiotherapy (Lujan et al. 1999) and a localization validation (Wittkampf et al. 1999).

What a reader should ask:

  • Were points gated, and on what signal and window? (Equation 2)
  • What does the reference catheter correct, and how well does it correlate with the site? (Equation 3)
  • Was drift checked by revisiting landmarks? (Equation 5)

Visual summary

Schematic

gate near exhale mapping pointreference subtract only if well correlated landmark visits: drift
Figure 1: Gating near exhale, a reference catheter, and drift from landmark visits.

Assumptions

How experts think

When a map looks blurred or shifted, ask:

  1. Which motion is it: breathing, heartbeat or drift? Each has its own time scale.
  2. Was acquisition gated, and how wide was the window? (Equation 2)
  3. Is the reference correlated with this site, and was it scaled? (Equation 3, Equation 4)
  4. When were landmarks last revisited? (Equation 5)
  5. How much lag does the smoothing add? (Equation 6)

One sentence to keep: every correction for motion trades something: time, lag or the risk of making it worse.

Cheat sheet

Quantity Result
Breathing (Equation 1) \(z = -b\cos^{2n}(\pi t/T)\)
Gating (Equation 2) \(D = 1 - (2/\pi)\arccos((\delta/b)^{1/(2n)})\); residual \(\le \delta\)
Reference (Equation 3) \(\sqrt{\sigma_x^2 + \sigma_r^2 - 2\rho\sigma_x\sigma_r}\); helps iff \(\rho > \sigma_r/(2\sigma_x)\)
Surrogate (Equation 4) \(\sigma_x\sqrt{1 - \rho^2}\) with \(k = \rho\sigma_x/\sigma_r\)
Drift (Equation 5) slope; SE \(= \sigma/\sqrt{\sum(t_i - \bar t)^2}\)
Averaging (Equation 6) \(\lvert\operatorname{sinc}(fW)\rvert\); delay \(W/2\)

Values used in this module

Value Source
22 patients; 7.8 ± 2.1 vs 3.2 ± 1.0 min Beinart et al. (2011)
Coronary–diaphragm slope about 0.6, range 0.17–0.93; \(r^2\) 0.04–0.87; 12 volunteers Danias et al. (1999)
Landmark reproducible within 1.4 ± 1.1 mm after 2 h Wittkampf et al. (1999)
Breathing period and excursion, correlations, drift rate, noise illustrative
Duty cycles, residuals, standard errors, 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. Why is respiratory gating usually done at end expiration?

Model answer

Breathing spends more time near exhale, so a window there accepts more time for the same residual motion (Equation 1, Equation 2) (Lujan et al. 1999).

Q2. Derive the gating duty cycle for the cos^{2n} model.

Model answer

\(\cos^{2n}(\pi t/T) \le \delta/b\) means \(\lvert\cos(\pi t/T)\rvert \le (\delta/b)^{1/(2n)}\); the accepted fraction of a period is \(1 - (2/\pi)\arccos\) of that (Equation 2).

Q3. What did respiratory compensation change in one clinical study?

Model answer

Ungated maps overestimated pulmonary vein ostia against CT or MRI; gated maps took longer to collect (Beinart et al. 2011).

Q4. When does subtracting a reference catheter make things worse?

Model answer

When \(\rho < \sigma_r/(2\sigma_x)\): the reference adds its own uncorrelated motion (Equation 3).

Q5. How can you use a surrogate that moves less than the heart?

Model answer

Fit a slope and subtract \(k r\); the residual is \(\sigma_x\sqrt{1 - \rho^2}\) (Equation 4).

Q6. Why fit the slope per patient?

Model answer

The relation between heart and surrogate motion varies widely between people, with slopes from 0.17 to 0.93 in one MR study (Danias et al. 1999).

Q7. How would you estimate impedance drift during a case?

Model answer

Revisit a landmark at intervals and fit a line; its standard error is \(\sigma/\sqrt{\sum(t_i - \bar t)^2}\) (Equation 5).

Q8. Why spread the visits out?

Model answer

The slope’s error falls as the spread of visit times grows (Equation 5).

Q9. What does averaging a catheter position over one breath do?

Model answer

Removes breathing completely and delays the display by half a breath (Equation 6).

Q10. Gating and averaging both reduce breathing motion. How do they differ?

Model answer

Gating keeps the instantaneous position but only at some times; averaging uses all times but shows a delayed, smoothed position (Equation 2, Equation 6).

Q11. A landmark shifted 3 mm over two hours. Is that drift or noise?

Model answer

Compare with the slope’s standard error from several visits, and with known reproducibility, such as 1.4 ± 1.1 mm at 2 hours in one study (Equation 5) (Wittkampf et al. 1999).

Q12. A gated map took twice as long. How would you argue for or against gating?

Model answer

Weigh residual motion against duty cycle (Equation 2) and the evidence that gating changed anatomical dimensions (Beinart et al. 2011); the answer depends on what the map is for.

Code lab

The lab builds motion corrections.

Part What you do Data
A A breathing model (Equation 1) synthetic
B Respiratory gating (Equation 2) synthetic
C Reference catheters and surrogates (Equation 3, Equation 4) synthetic
D Drift from repeated visits (Equation 5) synthetic
E Averaging and delay (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. \(n = 1\), window equal to half the excursion. Duty cycle?

Answer

\(1 - (2/\pi)\arccos(\sqrt{0.5}) = 1 - (2/\pi)(\pi/4) = 0.5\) (Equation 2).

S2. Reference moves as much as the point, \(\rho\) = 0.5. Does subtraction help?

Answer

No: it leaves the SD unchanged (Equation 3).

S3. \(\rho\) = 0.8. Residual SD after regression, as a fraction?

Answer

0.6 (Equation 4).

S4. Visit times spread twice as widely, same count. Drift standard error?

Answer

Halved (Equation 5).

S5. Breathing period 4 s, averaging window 4 s. Remaining breathing motion and delay?

Answer

None for sinusoidal breathing; 2 s (Equation 6).

S6. Gating window 2 mm. Largest motion among accepted points?

Answer

2 mm (Equation 2).

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.

  • Gated and ungated maps. Beinart and colleagues’ comparison of left atrial maps with and without respiratory compensation (Beinart et al. 2011). This module: gating.
  • Heart against diaphragm. Danias and colleagues’ coronary and diaphragm displacements during free breathing (Danias et al. 1999). This module: reference catheters and surrogates.

Further reading

  • Breathing motion. An asymmetric breathing model (Lujan et al. 1999); heart and diaphragm motion (Danias et al. 1999).
  • Mapping. Respiratory compensation in left atrial mapping (Beinart et al. 2011); landmark reproducibility (Wittkampf et al. 1999).
  • Related modules. m05 (time–frequency tools), m16 (magnetic localization), m17 (impedance localization), m20 (geometry), m21 (image integration).

References

Beinart, Roy, Rajesh Kabra, Kevin E Heist, et al. 2011. “Respiratory Compensation Improves the Accuracy of Electroanatomic Mapping of the Left Atrium and Pulmonary Veins During Atrial Fibrillation Ablation.” Journal of Interventional Cardiac Electrophysiology 32 (2): 105–10. https://doi.org/10.1007/s10840-011-9583-z.
Danias, P G, M Stuber, R M Botnar, K V Kissinger, R R Edelman, and W J Manning. 1999. “Relationship Between Motion of Coronary Arteries and Diaphragm During Free Breathing: Lessons from Real-Time MR Imaging.” AJR. American Journal of Roentgenology 172 (4): 1061–65. https://doi.org/10.2214/ajr.172.4.10587147.
Lujan, A E, E W Larsen, J M Balter, and R K Ten Haken. 1999. “A Method for Incorporating Organ Motion Due to Breathing into 3D Dose Calculations.” Medical Physics 26 (5): 715–20. https://doi.org/10.1118/1.598577.
Wittkampf, F H, E F Wever, R Derksen, et al. 1999. “LocaLisa: New Technique for Real-Time 3-Dimensional Localization of Regular Intracardiac Electrodes.” Circulation 99 (10): 1312–17. https://doi.org/10.1161/01.cir.99.10.1312.