Skip to main content

m50 · Intracardiac echocardiography with TEE and TTE: ultrasound essentials for ICE, phased-array steering, frame rate, transseptal guidance, evidence

Warning

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

In 60 seconds

Intracardiac echocardiography (ICE) puts a small phased-array ultrasound probe on a catheter inside the heart. Five ideas explain most of what it can and cannot show. Depth comes from echo time, using one assumed speed of sound, 1540 m/s (Zhang et al. 2026). Depth resolution is half the pulse length, so it improves with frequency. Signal is lost to attenuation in proportion to frequency and depth; a probe inside the heart is close to its targets, so it can use the high frequencies that resolve fine detail, and studies report ICE imaging at 7.5–10 MHz (Morton et al. 2003). The beam is steered by delays across the elements, and elements spaced too far apart create false copies of it. And each image line waits for its deepest echo, which caps the frame rate. In the EP lab ICE guides transseptal puncture with less fluoroscopy (Abdalla et al. 2026); against transesophageal echo (TEE), most comparisons are observational and their certainty is low (Kanniyan et al. 2026; Yaser et al. 2026).

From the field

  • Anatomy you cannot map. With ICE at 7.5–10 MHz, the cavotricuspid isthmus measured 35 ± 6 mm long at end systole and 2.7–4.1 mm thick before ablation, with a pouch or recess in 11 of 15 patients (Morton et al. 2003).
  • Transseptal puncture without fluoroscopy. In a randomized trial of 43 patients comparing two ICE-guided puncture strategies, every procedure was done without fluoroscopy and without complications; one strategy was faster at first (8.5 against 12.0 minutes), the other caught up after about 10 cases (Chen et al. 2026). Pooled across eight studies, ICE-guided puncture used 2.07 minutes less fluoroscopy than TEE- or fluoroscopy-guided puncture, with similar first-pass success and no increase in safety events (Abdalla et al. 2026).
  • ICE or TEE. For AF ablation, a meta-analysis of six studies (4,760 patients, one randomized) found no significant difference in complications and shorter fluoroscopy with ICE, with low to very low certainty (Kanniyan et al. 2026). For left atrial appendage occlusion, 27 observational studies showed similar success, shorter procedures in adjusted analyses, and more pericardial effusions needing intervention with ICE (Yaser et al. 2026).
  • Associations in large databases. In a US readmissions database, ICE was used in 15.7% of 51,129 AF ablations and was associated with fewer 90-day readmissions (hazard ratio 0.88) (Deshpande et al. 2022); in a Japanese database of VT ablations it was associated with less cardiac tamponade (Kitamura et al. 2021). With pulsed-field ablation, a multicenter cohort found no improvement in procedural metrics with ICE (Dello Russo et al. 2023).
  • Shells from ICE. One center compared a deep-learning module that builds the left atrial shell from ICE with its earlier manual contouring, in a before-and-after design (Zou et al. 2026) (m40).

Derivations

Echo time to depth

A pulse leaves the probe, reflects, and returns after time \(t\). With speed of sound \(c\) the reflector is at depth

\[ z = \frac{c\,t}{2}, \qquad \hat z = z\,\frac{c_{\text{assumed}}}{c_{\text{true}}} \tag{1}\]

where \(\hat z\) is the depth the scanner draws when it assumes \(c_{\text{assumed}}\) and sound actually travels at \(c_{\text{true}}\). Scanners commonly assume 1540 m/s (Zhang et al. 2026); tissue differs, for example calf muscle measured 1560–1586 m/s in vivo (Xiao et al. 2024). At 1586 m/s a 35-mm isthmus (Morton et al. 2003) would be drawn about 1.0 mm short (lab part A). The error is a fixed fraction of every distance, so it scales with depth.

Key takeaways

  • Depth is time times an assumed speed; a wrong speed scales every distance.
  • Tracked ICE contours placed in 3-D inherit that scale error.

Axial resolution

The wavelength is \(\lambda = c/f\) for frequency \(f\). A pulse of \(n\) cycles is \(n\lambda\) long, and two echoes stop overlapping when their reflectors are more than half that apart:

\[ \lambda = \frac{c}{f}, \qquad \Delta z_{\text{axial}} = \frac{n\,\lambda}{2} \tag{2}\]

At 7.5 MHz the wavelength is 0.205 mm, and a 2-cycle pulse (illustrative) separates reflectors 0.21 mm apart; at 10 MHz, 0.15 mm (lab part B). Both are much finer than the isthmus walls ICE measured (Morton et al. 2003). Resolution across the beam is set by the aperture instead (see steering).

Key takeaways

  • Shorter pulses and higher frequencies resolve thinner layers along the beam.
  • Resolution along the beam and across it are different quantities.

Attenuation and how deep an image reaches

Attenuation in soft tissue grows roughly in proportion to frequency. With attenuation slope \(\alpha\) in dB/(cm·MHz), a round trip to depth \(z\) costs

\[ L = 2\,\alpha\,f\,z \ \text{dB}, \qquad z_{\max} = \frac{B}{2\,\alpha\,f} \tag{3}\]

where \(z_{\max}\) is the depth at which the loss uses up a signal budget \(B\). Freshly excised myocardium measured 0.52 dB/(cm·MHz) across the fibers and 1.88 along them (Baldwin et al. 2007). With an illustrative 60 dB budget and the across-fiber value, the reach is 7.7 cm at 7.5 MHz and 5.8 cm at 10 MHz (lab part C). ICE can afford high frequencies because it sits close to its targets.

Key takeaways

  • Higher frequency buys resolution and costs depth.
  • Attenuation depends on tissue and even on fiber direction.

Steering a phased array

A phased array of \(N\) elements spaced \(d\) apart steers its beam to angle \(\theta_0\) by firing element \(k\) at position \(x_k\) with delay \(\tau_k\). Its response in direction \(\theta\) is the array factor:

\[ \tau_k = \frac{x_k \sin\theta_0}{c}, \qquad \mathrm{AF}(\theta) = \frac{1}{N}\left|\sum_{k=0}^{N-1} e^{\,i 2\pi k \frac{d}{\lambda}(\sin\theta - \sin\theta_0)}\right|, \qquad \sin\theta_m = \sin\theta_0 + m\frac{\lambda}{d} \tag{4}\]

The array factor is 1 at \(\theta_0\) and again at every grating-lobe angle \(\theta_m\) (\(m = \pm1, \pm2, \ldots\)) that exists; with \(d \le \lambda/2\) none exists for any steering angle. Its first null lies \(\lambda/(Nd)\) away in \(\sin\theta\), so a wider aperture gives a narrower beam. An echo arriving through a grating lobe is drawn in the steered direction: a ghost.

Exact

Assumptions

  • Model: Equation 4 for identical point elements in a line, in the far field, at a single frequency; the JavaScript is tested against epml.ice.
  • Element spacing is given in wavelengths, so the result holds at any frequency; the readout converts it to millimeters at 7.5 MHz (Morton et al. 2003). Element counts are illustrative, not the specification of any catheter.
  • What would change the result: element directivity, apodization (weighting the elements), wideband pulses (which smear grating lobes) and delay errors.

Key takeaways

  • Delays steer the beam; spacing above half a wavelength creates ghosts.
  • Lateral resolution comes from the aperture, not the pulse.

Frame rate

Each line must wait for echoes from the deepest point \(D\), a time \(2D/c\). A frame of \(N_{\text{lines}}\) lines therefore refreshes at

\[ \mathrm{FR} = \frac{c}{2\,D\,N_{\text{lines}}} \tag{5}\]

With 96 lines (illustrative), a 4-cm image refreshes about 200 times a second and a 12-cm image about 67 times (lab part E). Deeper, wider or denser images are slower.

Exact

Assumptions

  • Model: Equation 2, Equation 3 and Equation 5 with the assumed 1540 m/s (Zhang et al. 2026); the JavaScript is tested against epml.ice.
  • Attenuation slopes are the myocardial values of Baldwin et al. (2007), applied to the whole path; the path in ICE is mostly blood, which is not modeled separately. The frequency range follows Morton et al. (2003). The 60 dB budget, the 2-cycle pulse and the line counts are illustrative.
  • What would change the result: focusing, parallel receive beams (several lines per transmit), harmonic imaging and time-gain compensation, none of which are modeled.

Key takeaways

  • Depth, lines and frame rate trade against each other.
  • A deep, wide image of a fast-moving catheter is the hardest case.

From planes to a 3-D shell

When the mapping system tracks the ICE catheter, each imaging plane has a known pose, so contours drawn on it become 3-D points, and a shell can be built from them (m20). In lab part F, planes rotated about the catheter axis leave gaps that shrink as the plane count grows, while point accuracy is set by contour noise and, through Equation 1, by the speed of sound. A published comparison of automated and manual ICE shells is a before-and-after study at one center (Zou et al. 2026).

Key takeaways

  • More planes close the gaps; they do not make each point more accurate.
  • A shell is only as good as the tracking, the contours and the assumed speed of sound.

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 ICE shows up in EP. Studies describe ICE for transseptal puncture (Chen et al. 2026; Abdalla et al. 2026), for guiding AF ablation as an alternative to TEE (Kanniyan et al. 2026), for appendage occlusion (Yaser et al. 2026), for imaging anatomy such as the cavotricuspid isthmus (Morton et al. 2003), and in VT ablation (Kitamura et al. 2021).

What the evidence is. The comparisons of ICE with TEE are meta-analyses of mostly observational studies, and their authors call the certainty low; no significant difference there means imprecision, not demonstrated equivalence (Kanniyan et al. 2026). Database studies associate ICE with outcomes but cannot separate ICE from the centers and operators who use it (Deshpande et al. 2022; Kitamura et al. 2021). The one randomized trial cited here compares two ways of using ICE, not ICE against another modality (Chen et al. 2026).

What a reader should ask:

  • Compared with what? TEE, fluoroscopy alone, or another way of using ICE.
  • Which outcome? Fluoroscopy time, procedure time, complications or recurrence.
  • Who did it? Experience shapes ICE results; one trial reported a learning curve of about 10 cases (Chen et al. 2026).

Visual summary

Schematic

ICE catheter (right atrium) septum and fossa (tenting) z = c t / 2 Axial resolution: n λ / 2 (pulse length) Reach: B / (2 α f) (attenuation) Steering: delays; d > λ/2 gives ghosts Frame rate: c / (2 D N lines) Tracked planes → 3-D shell (m20)
Figure 1: ICE in one picture: depth from echo time, resolution from pulse length, reach from attenuation, steering from delays, and frame rate from depth and line count.

Assumptions

How experts think

Before trusting an ICE image or a number measured on it, ask:

  1. Is the depth right? It assumes 1540 m/s; real tissue differs (Zhang et al. 2026; Xiao et al. 2024) (Equation 1).
  2. Can it resolve this? Compare the structure with \(n\lambda/2\) along the beam and with the beam width across it (Equation 2, Equation 4).
  3. Is that echo real? Grating lobes, side lobes and reverberation draw echoes where nothing is (Equation 4).
  4. Is it fast enough? Depth and line count set the frame rate for a moving catheter (Equation 5).
  5. What does the evidence compare? ICE against TEE, against fluoroscopy, or ICE used two ways (Kanniyan et al. 2026; Chen et al. 2026).

One sentence to keep: an ultrasound image is time turned into distance with an assumed speed, through a beam whose shape you chose.

Cheat sheet

Quantity Result
Depth from echo time (Equation 1) \(z = ct/2\); drawn at \(z\,c_{\text{assumed}}/c_{\text{true}}\)
Axial resolution (Equation 2) \(\lambda = c/f\); \(\Delta z = n\lambda/2\)
Attenuation and reach (Equation 3) \(L = 2\alpha f z\) dB; \(z_{\max} = B/(2\alpha f)\)
Steering (Equation 4) \(\tau_k = x_k\sin\theta_0/c\); grating lobes at \(\sin\theta_0 + m\lambda/d\); none if \(d \le \lambda/2\); first null \(\lambda/(Nd)\)
Frame rate (Equation 5) \(c/(2DN_{\text{lines}})\)

Values used in this module

Value Source
Assumed speed of sound 1540 m/s Zhang et al. (2026)
Calf muscle 1560–1586 m/s in vivo (example of real tissue) Xiao et al. (2024)
ICE imaging frequencies 7.5–10 MHz Morton et al. (2003)
Myocardial attenuation 0.52 (across fibers) and 1.88 (along) dB/(cm·MHz) Baldwin et al. (2007)

Studies in this module

Study Comparison Design
Abdalla et al. (2026) ICE vs TEE or fluoroscopy for transseptal puncture Meta-analysis, randomized and observational
Kanniyan et al. (2026) ICE vs TEE for AF ablation Meta-analysis, 5 cohorts and 1 randomized trial
Yaser et al. (2026) ICE vs TEE for appendage occlusion Meta-analysis, observational
Deshpande et al. (2022) ICE vs none, AF ablation Administrative database
Kitamura et al. (2021) ICE vs none, VT ablation Administrative database
Dello Russo et al. (2023) ICE vs none, pulsed-field ablation Multicenter cohort
Chen et al. (2026) Two ICE-guided puncture strategies Randomized trial

Interview questions

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

Q1. How does an ultrasound system know how deep a reflector is?

Model answer

From the echo’s round-trip time and an assumed speed of sound: \(z = ct/2\) (Equation 1). Systems commonly assume 1540 m/s (Zhang et al. 2026). If the tissue is faster or slower, every depth is scaled by \(c_{\text{assumed}}/c_{\text{true}}\).

Q2. What sets axial resolution, and how would you improve it?

Model answer

Half the spatial pulse length, \(n\lambda/2\) (Equation 2). Shorter pulses (fewer cycles, wider bandwidth) or a higher frequency improve it; higher frequency costs penetration (Equation 3).

Q3. Why can a probe inside the heart use a higher frequency than a probe that must image from far away?

Model answer

Round-trip loss is \(2\alpha f z\) (Equation 3): for the same signal budget, a shorter path allows a proportionally higher frequency, which gives finer resolution. ICE sits close to its targets; ICE studies report 7.5–10 MHz (Morton et al. 2003).

Q4. What is a grating lobe, and how is it avoided?

Model answer

A full-strength copy of the main beam at \(\sin\theta_0 + m\lambda/d\) (Equation 4), which appears when the elements are spaced more than half a wavelength apart for the steering range used. Echoes from it are drawn in the wrong direction. Spacing elements at \(\lambda/2\) or less prevents it at any steering angle.

Q5. How do you narrow the beam of a phased array?

Model answer

Widen the aperture: the first null is \(\lambda/(Nd)\) from the main lobe in \(\sin\theta\) (Equation 4), so more elements, or a higher frequency for the same aperture, give a narrower beam. On a catheter the aperture is limited by its diameter.

Q6. Your ICE image updates too slowly to follow the catheter. What can you change?

Model answer

Reduce the depth or the number of lines (narrower sector or coarser line spacing), since the frame rate is \(c/(2DN_{\text{lines}})\) (Equation 5). Systems that form several receive lines per transmit can raise it further, at some cost in image quality.

Q7. A contour drawn on tracked ICE planes is merged into the map. List its error sources.

Model answer

Catheter tracking error, the speed-of-sound scale (Equation 1), contour placement (resolution and noise), cardiac and respiratory phase, and gaps between planes that must be interpolated (m20). More planes reduce the gaps but not the per-point error.

Q8. What does the evidence say about ICE against TEE for AF ablation?

Model answer

A meta-analysis of six studies, one randomized, found no significant difference in complications and shorter fluoroscopy with ICE, with low to very low certainty; the authors note that no difference reflects imprecision, not equivalence (Kanniyan et al. 2026).

Q9. A database study links ICE use to fewer readmissions. Does ICE cause that?

Model answer

Not shown. The association (hazard ratio 0.88 for 90-day readmission (Deshpande et al. 2022)) comes from administrative codes with propensity matching; centers and operators who use ICE may differ in ways the data do not capture. It generates a hypothesis for a randomized comparison.

Q10. Why might ICE help less with some catheters than others?

Model answer

Because its benefit depends on what else guides the procedure. In a multicenter pulsed-field ablation cohort, ICE did not improve procedural metrics (Dello Russo et al. 2023); when a catheter is already well visualized by the mapping system, ICE adds less to navigation and more to safety monitoring. That is a hypothesis to test, not a finding.

Q11. What does ICE show that an electroanatomic map cannot?

Model answer

Tissue it never touches: wall structure such as isthmus pouches and thickness (Morton et al. 2003), and the fossa during puncture (Chen et al. 2026). The map only knows where the catheter has been and what it recorded.

Q12. A structure looks 1 mm shorter on ICE than on CT. Name a physical reason.

Model answer

The speed of sound: if the tissue is faster than the assumed 1540 m/s, every distance is drawn shorter in proportion (Equation 1). At 1586 m/s a 35-mm structure is drawn about 1 mm short. Other reasons include the cardiac phase of each image and an oblique imaging plane.

Code lab

The lab builds the ultrasound essentials from first principles and ends with the published evidence.

Part What you do Data
A Echo time to depth, and the speed-of-sound assumption (Equation 1) published values
B When two echoes stop overlapping (Equation 2) our own pulses
C Attenuation and how deep an image reaches (Equation 3) published attenuation slopes
D Steering, main lobe and grating lobes (Equation 4) our own array
E A sector image of the septum, tenting, and frame rate (Equation 5) our own phantom
F A 3-D shell from tracked 2-D planes our own phantom
G Reading the ICE evidence by design 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. An echo returns after 26 µs. How deep is the reflector at 1540 m/s?

Answer

\(1540 \times 26\times10^{-6}/2 \approx 0.020\) m, or 20 mm (Equation 1).

S2. What is the wavelength at 7.7 MHz, assuming 1540 m/s?

Answer

\(1540/7.7\times10^{6} = 0.2\) mm (Equation 2).

S3. With \(\alpha\) = 0.5 dB/(cm·MHz) at 10 MHz, what does a round trip to 3 cm cost?

Answer

\(2 \times 0.5 \times 10 \times 3 = 30\) dB (Equation 3).

S4. Elements are spaced one wavelength apart and the beam points straight ahead. Where are the grating lobes?

Answer

At \(\sin\theta = \pm 1\), that is ±90° (Equation 4): at the edge of the field, and they move inward as soon as the beam is steered.

S5. Doubling the image depth at the same line count does what to the frame rate?

Answer

Halves it (Equation 5).

S6. True or false: adding more tracked planes makes each ICE contour point more accurate.

Answer

False. More planes close the gaps between them; per-point accuracy is set by tracking, contour noise and the speed of sound (lab part F).

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.

  • ICE-guided transseptal puncture, pooled. Abdalla and colleagues pool fluoroscopy time, radiation dose and safety across eight studies [Abdalla et al. (2026); free full text]. This module: clinical correlate.
  • Two puncture strategies. Chen and colleagues compare two ICE-guided strategies and their learning curves [Chen et al. (2026); free full text]. This module: from the field.
  • ICE or TEE for appendage occlusion. Yaser and colleagues report adjusted and unadjusted pooled estimates [Yaser et al. (2026); free full text]. This module: clinical correlate.

Further reading

References

Abdalla, Asim Mohammed Eldai, Kolluru Pavani Durga, Animisha Chukka, et al. 2026. “Comparative Outcomes of ICE Versus TEE/Fluoroscopy Guidance Transseptal Puncture in LA Procedures: A Systematic Review and Meta-Analysis.” Annals of Noninvasive Electrocardiology 31 (3): e70186. https://doi.org/10.1111/anec.70186.
Baldwin, Steven L, Min Yang, Karen R Marutyan, et al. 2007. “Ultrasonic Detection of the Anisotropy of Protein Cross Linking in Myocardium at Diagnostic Frequencies.” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 54 (7): 1360–69. https://doi.org/10.1109/tuffc.2007.396.
Chen, Chun-Kai, An-Li Yu, Su-Huan Chang, et al. 2026. “Comparison of Two ICE-Guided Transseptal Puncture Strategies for Zero-Fluoroscopy AF Ablation: A Randomized Controlled Trial.” Journal of Arrhythmia 42 (2): e70322. https://doi.org/10.1002/joa3.70322.
Dello Russo, A, C Tondo, V Schillaci, et al. 2023. “Intracardiac Echocardiography-Guided Pulsed-Field Ablation for Successful Ablation of Atrial Fibrillation: A Propensity-Matched Analysis from a Large Nationwide Multicenter Experience.” Journal of Interventional Cardiac Electrophysiology 67 (5): 1257–66. https://doi.org/10.1007/s10840-023-01699-2.
Deshpande, S, H Sawatari, R Ahmed, et al. 2022. “Impact of Intracardiac Echocardiography on Readmission Morbidity and Mortality Following Atrial Fibrillation Ablation.” Journal of Cardiovascular Electrophysiology 33 (12): 2496–503. https://doi.org/10.1111/jce.15683.
Kanniyan, Afthab Salam, Omer Mohammed, Arshita Wadhwa, et al. 2026. “Intracardiac Echocardiography Versus Transesophageal Echocardiography for Guidance of Atrial Fibrillation Ablation: A Systematic Review and Meta-Analysis of Procedural Safety and Efficacy.” Echocardiography 43 (7): e70543. https://doi.org/10.1111/echo.70543.
Kitamura, T, M Nakajima, I Kawamura, et al. 2021. “Safety and Effectiveness of Intracardiac Echocardiography in Ventricular Tachycardia Ablation: A Nationwide Observational Study.” Heart and Vessels 36 (7): 1009–15. https://doi.org/10.1007/s00380-020-01766-y.
Morton, Joseph B, Prashanthan Sanders, Neil C Davidson, et al. 2003. “Phased-Array Intracardiac Echocardiography for Defining Cavotricuspid Isthmus Anatomy During Radiofrequency Ablation of Typical Atrial Flutter.” Journal of Cardiovascular Electrophysiology 14 (6): 591–97. https://doi.org/10.1046/j.1540-8167.2003.02152.x.
Xiao, Di, Pat De la Torre, and Alfred C H Yu. 2024. “Real-Time Speed-of-Sound Estimation in Vivo via Steered Plane Wave Ultrasound.” IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 71 (6): 673–86. https://doi.org/10.1109/TUFFC.2024.3395490.
Yaser, Hatem, Shehab Yaser, Hazem E Mohammed, et al. 2026. “Intracardiac Versus Transesophageal Echocardiography for Left Atrial Appendage Occlusion: A Meta-Analysis.” JACC: Advances 5 (6 Pt 2): 102815. https://doi.org/10.1016/j.jacadv.2026.102815.
Zhang, Jiaxin, Yunlong Zhu, and Muyinatu A Lediju Bell. 2026. “Spatial Coherence-Based Sound Speed Estimation in Plane Wave Ultrasound Imaging.” IEEE Transactions on Ultrasonics 73 (1): 57–69. https://doi.org/10.1109/TUSON.2025.3633606.
Zou, Fengwei, Sarah Xu, Sanjana Nagraj, et al. 2026. “Atrial Fibrillation Ablation Using Three-Dimensional Artificial Intelligence Module Integration with Intracardiac Echocardiography.” Europace 28 (2). https://doi.org/10.1093/europace/euag009.