Skip to main content

m53 · Imaging ablation lesions and real-time MRI-guided EP: lesions, gaps and edema, interventional MRI and its limits

Warning

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

In 60 seconds

After ablation, late gadolinium enhancement (LGE) and edema-sensitive MRI show the treated tissue (Fusco et al. 2024). Whether they show a gap depends on the gap’s width against the image’s blur: a narrow gap in a blurred line fills in. Three months after pulmonary vein isolation, the number of scar gaps on MRI tracked how many veins had reconnected, but their location agreed only fairly with where veins reconnected (kappa 0.37) (Al Jefairi et al. 2019); in a later trial, lesions did not uniformly become scar (Akoum et al. 2024). Right after ablation, edema makes lesions look larger than the irreversible injury (Tampakis et al. 2023). Real-time MRI-guided ablation avoids ionizing radiation; in one early series of typical flutter, acute block was confirmed in 32 of 35 patients and technical challenges occurred in 19 (Hopman et al. 2026). Its limits are compatible equipment, electrograms recorded under strong interference, catheter visualization and lesion assessment (Tampakis et al. 2023).

From the field

  • Gaps predict how many, not where. In 51 patients scanned three months after isolation and then studied electrically, scar gaps were found in 76% of patients and 39% of veins, and reconnection in 88% of patients and 50% of veins. The number of gaps correlated with the extent of reconnection, but location agreement was fair (kappa 0.37), and gaps were missed in areas of pre-existing fibrosis (Al Jefairi et al. 2019).
  • Edema is not scar. Edema seen on T2-weighted imaging after ablation includes reversible injury and is larger than the delayed-enhancement region; it subsides over time, and electrical reconnection can follow (Tampakis et al. 2023).
  • Isolation despite a gap. Electrical isolation can be observed even where MRI shows a gap in the tissue (Tampakis et al. 2023), so an image gap and a reconnected vein answer different questions.
  • Real-time MRI in early use. In a single-center series of 35 typical-flutter ablations at 1.5 T, acute bidirectional block was confirmed in 32; challenges, most often with coronary sinus catheter positioning and system integration, occurred in 19; typical flutter recurred in 6 of 32 within a year. The authors describe an early implementation phase of a first-generation platform (Hopman et al. 2026).
  • Fast anatomy for guidance. In 42 patients, a 3-D whole-heart scan took about 98 s and automatic meshes of the right-sided chambers took about 16 s; ablation followed the predefined line in 62% (Jahnke et al. 2024).
  • Frames and tracking. Real-time images for passive catheter visualization have been reported at 4–8 frames per second, and active tracking of tip coils at up to 50 (Tampakis et al. 2023).

Derivations

When blur hides a gap

Take a straight lesion line of width \(w\) with a gap of width \(g\) along it, and an image blurred by an isotropic Gaussian of standard deviation \(\sigma\). The 2-D Gaussian is separable, so the blurred image is the blurred band across the line times one minus the blurred gap along it:

\[ I(x, y) = B(y)\,\bigl[1 - G(x)\bigr], \qquad I_{\text{line}} = \operatorname{erf}\!\left(\frac{w}{2\sqrt2\,\sigma}\right), \qquad I_{\text{gap}} = I_{\text{line}}\left[1 - \operatorname{erf}\!\left(\frac{g}{2\sqrt2\,\sigma}\right)\right] \tag{1}\]

where \(B\) and \(G\) are blurred box profiles, the same error functions as a thin wall in m51 (eq-m51-wall). With a threshold \(\theta\) (here half the contrast), the gap is visible when \(I_{\text{gap}} < \theta\), that is when \(g > 2\sqrt2\,\sigma\,\operatorname{erfinv}(1 - \theta/I_{\text{line}})\). LGE resolution is about 1.5–2 mm routinely and about 1 mm when thin walls are the target (Kellman and Arai 2012); for a 4-mm line (illustrative), the smallest visible gap is 0.57, 0.86 and 1.12 mm at 1, 1.5 and 2 mm blur (lab part A). If acute edema extends a margin \(e\) beyond each lesion, a gap appears as \(\max(0, g - 2e)\) (lab part D).

Exact

Assumptions

  • Model: Equation 1 for a straight band with a rectangular gap and an isotropic Gaussian blur; tested against epml.lesion.
  • The visibility threshold is half the full contrast. The blur range covers the published LGE resolutions (Kellman and Arai 2012); line and gap widths are illustrative.
  • What it ignores: noise, curvature of real lines, partial volume across the wall’s thickness, edema, and lesions whose signal is not uniform.

Key takeaways

  • A gap narrower than about the blur width fills in.
  • A thin or faint line can vanish with its gaps.

Gaps against reconnection

An image gap is anatomy; reconnection is physiology. Agreement between them per vein or per segment is measured with Cohen’s kappa (m35). In a simulation where gaps 1 mm or wider usually conduct and veins also reconnect for other reasons, even a perfect image of every gap reached a kappa of only 0.63, and blurring away gaps too thin to conduct could raise the agreement slightly (lab part C; all probabilities illustrative). The clinical figure, 0.37 for location (Al Jefairi et al. 2019), mixes these effects with registration, pre-existing fibrosis and timing.

Key takeaways

  • Kappa measures agreement between two different questions, not the scanner alone.
  • Thin gaps, recovery and pre-existing fibrosis all separate the image from the electrogram.

Real-time MRI: frames and tracking

A real-time image is acquired line by line, one repetition time TR apart. With \(N_{\text{lines}}\) lines and an acceleration factor \(R\),

\[ T_{\text{frame}} = \frac{N_{\text{lines}}\,\mathrm{TR}}{R}, \qquad \text{frame rate} = \frac{1}{T_{\text{frame}}}, \qquad d = v\,T_{\text{frame}} \tag{2}\]

where \(d\) is how far a structure moving at speed \(v\) travels during one frame. With 64 lines and an illustrative TR of 3 ms, a frame takes 192 ms, about 5 frames per second, inside the 4–8 reported for passive catheter visualization; active tracking of a few tip coils has been reported at up to 50 updates per second (Tampakis et al. 2023). For scale, coronary arteries moved 22–70 mm/s in plane (Achenbach et al. 2000), so 9 mm per 192-ms frame at their average speed.

Exact

Assumptions

  • Model: Equation 2 for single-shot line-by-line acquisition; tested against epml.lesion.
  • Reported rates are from a review (Tampakis et al. 2023); speeds are the published coronary averages (Achenbach et al. 2000). Repetition times, line counts and acceleration factors are illustrative.
  • What it ignores: non-Cartesian and partial acquisitions, view sharing, reconstruction latency and the time to display and act on an image.

Key takeaways

  • Fewer lines or more acceleration buy frames at the cost of resolution or signal.
  • Tracking a point is much faster than imaging a plane.

Electrograms inside the scanner

Electrograms recorded during scanning carry magnetohydrodynamic voltages from flowing blood, RF pickup and voltages induced by gradient switching. One setup combined RF filtering with a 300 Hz low-pass, a 30 Hz high-pass and a 60 Hz notch, and obtained the best signals with a 120 Hz low-pass despite lower amplitude (Tampakis et al. 2023). In lab part F, with illustrative waveforms, the 120 Hz low-pass left about a third of the residual interference of 300 Hz but shrank the electrogram from 1.64 to 1.48 mV peak to peak. Filters that remove interference also remove signal, which matters most for low-amplitude potentials (m03, m14).

Key takeaways

  • Scanner interference overlaps the electrogram’s band, so filtering always trades signal for noise.
  • Interpretation of low-amplitude signals is the first casualty.

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 lesion imaging shows up in EP. Studies image lesions after pulmonary vein isolation and compare gaps with reconnection at a repeat procedure (Al Jefairi et al. 2019), examine whether lesions became scar in a trial (Akoum et al. 2024), and describe edema and enhancement after ablation (Fusco et al. 2024; Tampakis et al. 2023). Real-time MRI-guided ablation has been reported in typical atrial flutter (Hopman et al. 2026; Jahnke et al. 2024; Tampakis et al. 2023).

What the evidence is. Lesion-gap studies are single-center cohorts with a repeat electrophysiological study as the reference (Al Jefairi et al. 2019). The real-time MRI reports are single-center series and reviews of feasibility, and in humans the cited reports are all in typical flutter (Hopman et al. 2026; Jahnke et al. 2024; Tampakis et al. 2023). None is a randomized comparison with conventional guidance.

What a reader should ask:

  • When was the scan? Acute edema, early enhancement and chronic scar show different things (Tampakis et al. 2023).
  • How wide must a gap be to be seen? Compare it with the resolution (Equation 1).
  • Agreement with what, at what level? Per patient, per vein or per segment changes the answer (Al Jefairi et al. 2019).

Visual summary

Schematic

vein wide gap seen, narrow gap fills in gap g looks like g − 2e with edema frame = N lines × TR / R tip tracking: fewer lines, faster
Figure 1: Lesion imaging and real-time MRI in one picture: blur decides which gaps show, edema narrows them for a while, and frame time is lines times repetition time.

Assumptions

How experts think

Before trusting an image of an ablation line, ask:

  1. Is the gap wider than the blur? If not, the image cannot show it (Equation 1).
  2. Is this edema or scar? Acute imaging overstates the lesion (Tampakis et al. 2023).
  3. Was pre-existing fibrosis there? It hides gaps on LGE (Al Jefairi et al. 2019).
  4. Is the agreement per vein or per segment? Extent and location can disagree (Al Jefairi et al. 2019).
  5. For real-time guidance, how fast is the loop? Frame time and tracking rate set what the operator sees (Equation 2).

One sentence to keep: an image gap is anatomy seen through blur and edema at one moment; reconnection is conduction measured later.

Cheat sheet

Quantity Result
Gap in a blurred line (Equation 1) line \(\operatorname{erf}(w/(2\sqrt2\sigma))\); gap center line \(\times\,[1 - \operatorname{erf}(g/(2\sqrt2\sigma))]\); visible if \(g > 2\sqrt2\sigma\,\operatorname{erfinv}(1 - \theta/I_{\text{line}})\)
Edema margin \(e\) apparent gap \(\max(0, g - 2e)\)
Real-time frame (Equation 2) \(T = N\,\mathrm{TR}/R\); motion per frame \(vT\)
Agreement Cohen’s kappa (m35)

Values used in this module

Value Source
LGE resolution 1.5–2 mm routinely, about 1 mm for thin walls Kellman and Arai (2012)
4–8 frames/s passive visualization; up to 50/s active tracking Tampakis et al. (2023)
Filter chain 300 Hz low-pass, 30 Hz high-pass, 60 Hz notch; best at 120 Hz low-pass Tampakis et al. (2023)
Coronary in-plane speed 22.4–69.5 mm/s Achenbach et al. (2000)

Studies in this module

Study Question Design
Al Jefairi et al. (2019) Scar gaps against vein reconnection Prospective cohort, 51 patients
Akoum et al. (2024) Lesions and scar in a trial Secondary analysis of a randomized trial
Hopman et al. (2026) Real-time MRI-guided flutter ablation Retrospective single-center series, 35 patients
Jahnke et al. (2024) Meshes for MRI-guided ablation Series, 42 patients
Tampakis et al. (2023) State of real-time MRI-guided ablation Narrative review
Fusco et al. (2024) MRI before, during and after ablation Primer for radiologists

Interview questions

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

Q1. Derive the signal at the center of a gap in a blurred lesion line.

Model answer

The line minus the gap is a band times one minus a box; a separable Gaussian blur gives \(B(y)[1 - G(x)]\), so at the center it is \(\operatorname{erf}(w/(2\sqrt2\sigma))\,[1 - \operatorname{erf}(g/(2\sqrt2\sigma))]\) (Equation 1).

Q2. Your LGE resolution is 2 mm. Can you see a 0.5-mm gap?

Model answer

Not with a half-contrast threshold: for a 4-mm line the smallest visible gap at 2-mm blur is about 1.1 mm (Equation 1). A narrower gap reads as a continuous line.

Q3. Why might an MRI taken the day after ablation show no gaps, yet a vein reconnects?

Model answer

Acute edema makes lesions look wider than the irreversible injury, narrowing or closing gaps on the image; as it resolves, tissue can recover and conduct (Tampakis et al. 2023). A gap narrower than the blur is also invisible.

Q4. Image gaps predicted the number of reconnected veins but not their location. How can both be true?

Model answer

A patient with more lesion damage overall has more of both, so counts correlate; but placing a gap on the right segment needs resolution, registration and correct handling of pre-existing fibrosis, where gaps were missed (Al Jefairi et al. 2019). Location agreement was fair (kappa 0.37).

Q5. Why use kappa rather than percent agreement?

Model answer

Because two findings that are each common agree often by chance; kappa subtracts the agreement expected by chance (m35).

Q6. A real-time sequence uses 96 lines at a TR of 2.5 ms with no acceleration. What is the frame rate?

Model answer

\(96 \times 2.5\) = 240 ms per frame, about 4.2 frames per second (Equation 2): at the low end of the reported 4–8 for passive visualization (Tampakis et al. 2023).

Q7. Why can active tracking update so much faster than imaging?

Model answer

It locates a few small coils at the catheter tip from short projections instead of acquiring every line of a plane, so each update needs far less data; rates up to 50 per second have been reported (Tampakis et al. 2023).

Q8. What disturbs an electrogram inside an MRI scanner, and what does filtering cost?

Model answer

Magnetohydrodynamic voltages from flowing blood, RF pickup and gradient-induced voltages (Tampakis et al. 2023). Filters that remove them also remove part of the electrogram; one setup found the best signals at a 120 Hz low-pass despite lower amplitude. Low-amplitude potentials suffer first.

Q9. What would you need before using MRI lesion imaging as a procedural endpoint?

Model answer

Evidence that the acute image predicts durable lesions and outcome, a resolution that shows the gaps that matter (Equation 1), a sequence that separates edema from necrosis and can be repeated, and a prospective comparison with the current endpoint. The cited work does not yet provide this (Tampakis et al. 2023; Al Jefairi et al. 2019).

Q10. What did the early real-time MRI flutter series report, and how should it be read?

Model answer

Acute block in 32 of 35, challenges in 19, no major complications and recurrence in 6 of 32 within a year (Hopman et al. 2026). It is a retrospective single-center series of a first-generation platform: evidence of feasibility, not of superiority.

Q11. What did a lesion analysis inside a randomized trial add?

Model answer

That ablation lesions did not uniformly become scar on follow-up MRI, with large differences between centers (Akoum et al. 2024): delivery and measurement both vary.

Q12. You blur away thin gaps and agreement with reconnection goes up. Is the scanner better?

Model answer

No. Thin gaps that do not conduct were disagreements; hiding them raises kappa without making the image more accurate (lab part C). Agreement depends on what the reference measures.

Code lab

The lab goes from a single gap to a ring of lesions, then to real-time imaging and electrograms in the scanner.

Part What you do Data
A When blur hides a gap (Equation 1) published resolutions
B Gaps in a ring lesion around a vein our own phantom
C Image gaps against reconnection (kappa, m35) our own simulation
D Acute edema and apparent gaps our own model
E Real-time MRI frame budget (Equation 2) published frame rates
F Electrograms during scanning published filter chain, our own signals
G Reading the evidence published results

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. At the center of a very wide gap, what is the signal?

Answer

Zero: \(\operatorname{erf}(g/(2\sqrt2\sigma)) \to 1\) (Equation 1).

S2. A gap of 3 mm with an edema margin of 1 mm on each side looks how wide?

Answer

1 mm.

S3. 128 lines at TR 3 ms with acceleration 2: frame time?

Answer

192 ms (Equation 2).

S4. A tracked tip updates 50 times a second. How long between updates?

Answer

20 ms.

S5. Two raters agree on 90% of veins. Is their kappa 0.9?

Answer

Not necessarily: kappa subtracts chance agreement, which is high when one answer is common (m35).

S6. True or false: a lower low-pass cut-off always gives a better electrogram.

Answer

False. It removes more interference and more of the signal; the best setting depends on what must be seen (Tampakis et al. 2023).

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.

  • Real-time MRI-guided ablation. Tampakis and colleagues show an interventional MRI suite, electrograms disturbed by gradients, and images of flutter lesions before and after ablation, with edema and enhancement [Tampakis et al. (2023); free full text]. This module: real-time MRI and electrograms inside the scanner.
  • Meshes for guidance. Jahnke and colleagues report the whole-heart scan, the automatic segmentation and meshes used for navigation, and how closely ablation followed the planned line [Jahnke et al. (2024); free full text]. This module: from the field.

Further reading

References

Achenbach, S., D. Ropers, J. Holle, et al. 2000. “In-Plane Coronary Arterial Motion Velocity: Measurement with Electron-Beam CT.” Radiology 216 (2): 457–63. https://doi.org/10.1148/radiology.216.2.r00au19457.
Akoum, Nazem, Mario Mekhael, Felipe Bisbal, et al. 2024. “Lesion Delivery and Scar Formation in Catheter Ablation for Atrial Fibrillation: The DECAAF II Trial.” Heart Rhythm 22 (6): 1384–93. https://doi.org/10.1016/j.hrthm.2024.08.062.
Al Jefairi, Nora, Claudia Camaioni, Soumaya Sridi, et al. 2019. “Relationship Between Atrial Scar on Cardiac Magnetic Resonance and Pulmonary Vein Reconnection After Catheter Ablation for Paroxysmal Atrial Fibrillation.” Journal of Cardiovascular Electrophysiology 30 (5): 727–40. https://doi.org/10.1111/jce.13908.
Fusco, Armando, Domenico De Santis, Ermenegildo De Ruvo, et al. 2024. “Radiofrequency Ablation Guided by Real-Time Cardiovascular Magnetic Resonance.” La Radiologia Medica 129 (12): 1823–29. https://doi.org/10.1007/s11547-024-01911-x.
Hopman, Luuk H. G. A., Sarah Mahmoodi, Raschel D. van Luijk, et al. 2026. “Interventional Cardiac Magnetic Resonance-Guided Flutter Ablation: Technical Evolution and Procedural Insights from Early Clinical Use.” Heart Rhythm 23 (11): e2789–96. https://doi.org/10.1016/j.hrthm.2026.07.024.
Jahnke, Cosima, Angeliki Darma, Frank Lindemann, et al. 2024. “Electrophysiological Cardiovascular MR: Procedure-Ready Mesh Model Generation for Interventional Guidance Based on Non-Selective Excitation Compressed Sensing Whole Heart Imaging.” Scientific Reports 14 (1): 8974. https://doi.org/10.1038/s41598-024-59230-0.
Kellman, Peter, and Andrew E. Arai. 2012. “Cardiac Imaging Techniques for Physicians: Late Enhancement.” Journal of Magnetic Resonance Imaging 36 (3): 529–42. https://doi.org/10.1002/jmri.23605.
Tampakis, Konstantinos, Sokratis Pastromas, Alexandros Sykiotis, et al. 2023. “Real-Time Cardiovascular Magnetic Resonance-Guided Radiofrequency Ablation: A Comprehensive Review.” World Journal of Cardiology 15 (9): 415–26. https://doi.org/10.4330/wjc.v15.i9.415.