Skip to main content

m51 · Cardiac CT for EP: attenuation and Hounsfield units, reconstruction, ECG gating, anatomy, thrombus, esophagus and wall thickness

Warning

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

In 60 seconds

Computed tomography (CT) sends X-rays through the body along many lines and measures how much of each beam is lost. Each measurement is a line integral of the attenuation coefficient; reconstruction turns a full set of them, a sinogram, into slices. Hounsfield units (HU) put water at 0 and air at −1000. At the 60 keV used in this module, blood and myocardium differ by about 5 HU, so the chambers are filled with iodinated contrast first. The heart moves, so cardiac CT keeps only data from a short window of the cardiac cycle, about half a gantry rotation with one X-ray source and a quarter with two (Petersilka et al. 2008; Mergen et al. 2023). Before ablation, CT shows atrial and venous anatomy, thrombus in the appendage (Romero et al. 2013), the course of the esophagus (Tsao et al. 2005) and wall thickness (Beinart et al. 2011); in ventricular tachycardia (VT), wall thinning has been studied as a marker of isthmuses (Ciaccio et al. 2024; Moore et al. 2024). Merged with the map, a surface fit reported 2.3 mm of error (Kistler et al. 2006), and in dogs, ablation at CT-marked targets missed by 1.8–2.7 mm (Dong et al. 2006).

From the field

  • Thrombus and predictive value. Against transesophageal echo (TEE), CT detected appendage thrombus with 96% sensitivity and 92% specificity across 19 studies, but at a pooled prevalence of 8.9% only 41% of positive scans were true thrombus. With a delayed scan the figures were 100%, 99% and 92% (Romero et al. 2013). Prevalence matters as much as the scanner (m35, m48).
  • Where the esophagus runs. In 48 patients with AF, the esophagus touched the middle of the posterior left atrium in every one. In 42 its lower part ran close to the left inferior pulmonary vein (mean shortest distance 2.8 mm), in 6 close to the right inferior vein (Tsao et al. 2005).
  • Walls thinner than the blur. Left atrial wall thickness on CT averaged 1.89 mm and ranged from 0.5 to 3.5 mm (Beinart et al. 2011): the same order as the blur of the image, so how it is measured changes the answer (lab parts F and G).
  • Wall thickness and VT isthmuses. In 61 patients with repaired tetralogy of Fallot (a congenital heart defect), median wall thickness at anatomical isthmuses was 2.6 mm where conduction was blocked, 4.8 mm where it was slow and 5.6 mm where it was normal (Moore et al. 2024). In 6 patients with VT after infarction, the wall near the isthmus was a few hundred micrometers thick, and regions with the largest relative change in thickness partly coincided with the isthmus boundaries (Ciaccio et al. 2024). Both are small single-center studies.
  • Faster windows, sharper images. In 30 patients scanned at a 0.25-s rotation, reconstructions with a 66-ms window had fewer motion artifacts and sharper vessels than 125-ms ones (Mergen et al. 2023).
  • Reported error is not target error. A surface fit to the map reported 2.3 ± 0.4 mm against 6.4 ± 2.8 mm for landmarks (Kistler et al. 2006); that is a distance between surfaces. In dogs with markers sewn to the heart, the error at the targets themselves was 1.8–2.7 mm by chamber (Dong et al. 2006).

Derivations

Attenuation and Hounsfield units

A beam of intensity \(I_0\) that crosses tissue with linear attenuation coefficient \(\mu\) leaves with \(I = I_0 e^{-\int \mu\,ds}\) (m49). So one detector reading gives a line integral, \(p = -\ln(I/I_0) = \int \mu\,ds\). CT reports \(\mu\) on a scale fixed by water (\(\mu_w\)) and air (\(\mu_a\)):

\[ \mathrm{HU} = 1000\,\frac{\mu - \mu_w}{\mu_w - \mu_a} \tag{1}\]

Water is 0 and air is −1000 by construction; we take \(\mu_a = 0\). With NIST mass attenuation coefficients at 60 keV (Hubbell and Seltzer 1995) (the same energy as m49), soft tissue is 54 HU, blood 59, fat −89 and cortical bone 1935 (lab part A). Real CT beams contain many energies, so clinical values differ; the scale and the reasoning do not.

Blood and myocardium are 5 HU apart, too close to separate in a noisy image. Iodine attenuates far more (7.579 cm²/g at 60 keV against 0.2059 for water (Hubbell and Seltzer 1995)), and attenuation adds by mass. A concentration \(c\) in mg/mL raises the value by

\[ \Delta\mathrm{HU} = 1000\,\frac{c \cdot 10^{-3}\,(\mu/\rho)_{\mathrm I}}{\mu_w} \tag{2}\]

with \((\mu/\rho)_{\mathrm I}\) in cm²/g and \(\mu_w\) in 1/cm: 36.8 HU per mg/mL at 60 keV. The blood-pool concentration in our phantom, 10 mg/mL, is illustrative.

Key takeaways

  • A CT number is attenuation on a water-and-air scale; it depends on the beam energy.
  • Contrast is what separates blood from myocardium.

Projections and the central-slice theorem

Rotate the source and detector to angle \(\theta\). The detector at offset \(s\) records the integral of the image \(f(x, y)\) along the line \(x\cos\theta + y\sin\theta = s\). The set of all such readings is the Radon transform (Radon 1986), drawn as a sinogram:

\[ p_\theta(s) = \iint f(x, y)\,\delta(x\cos\theta + y\sin\theta - s)\,dx\,dy \tag{3}\]

For an ellipse of value \(v\) the integral is \(v\) times the chord, which has a closed form, so our phantom of ellipses has an exact sinogram (lab part B). Take the 1-D Fourier transform of one projection and substitute the line:

\[ P_\theta(k) = \int p_\theta(s)\,e^{-i2\pi k s}\,ds = \iint f(x, y)\,e^{-i2\pi k (x\cos\theta + y\sin\theta)}\,dx\,dy = F(k\cos\theta,\,k\sin\theta) \tag{4}\]

This is the central-slice theorem: each projection gives the 2-D spectrum \(F\) of the image along one line through the origin. Enough angles fill the spectrum, and the image follows.

Key takeaways

  • A sinogram is a stack of line integrals, one row per angle.
  • Each row fills one line of the image’s 2-D spectrum.

Filtered back-projection

Write the inverse 2-D Fourier transform in polar coordinates. The area element \(|k|\,dk\,d\theta\) brings in a factor \(|k|\), and with Equation 4:

\[ f(x, y) = \int_0^{\pi} q_\theta(x\cos\theta + y\sin\theta)\,d\theta, \qquad q_\theta = \mathcal{F}^{-1}\{\,|k|\,P_\theta(k)\,\} \tag{5}\]

Filter each projection with the ramp \(|k|\), then smear it back along its lines and add up the angles. Without the filter, back-projection over-weights low frequencies and blurs the image. Sampled every \(d\) mm and limited to the detector’s Nyquist frequency \(1/(2d)\), the ramp becomes the Ram-Lak kernel (Kak and Slaney 2001):

\[ h[0] = \frac{1}{4d^2}, \qquad h[n] = -\frac{1}{\pi^2 n^2 d^2}\ \ (n \text{ odd}), \qquad h[n] = 0\ \ (n \text{ even},\ n \neq 0) \tag{6}\]

It follows from \(2\int_0^{1/(2d)} k\cos(2\pi k n d)\,dk\), which tests/test_m51_ct.py checks symbolically. With 180 angles our reconstruction puts the left atrium at 430 HU against a true 427 (lab part C).

Simulation

Assumptions

  • Model: exact parallel-beam projections of an ellipse phantom (Equation 3), ramp-filtered with the Ram-Lak kernel (Equation 6) and back-projected with linear interpolation (Equation 5); our own code, tested against epml.ct. Grid 160 × 160 pixels of 2 mm; one detector bin per pixel; 6–180 angles over 180°.
  • Attenuation at one energy, 60 keV (Hubbell and Seltzer 1995); lungs are drawn as air, and the iodine concentration (10 mg/mL), photon counts and all shapes are illustrative. Noise uses a normal approximation to Poisson counts (Equation 7).
  • What would change the result: fan or cone beams, polyenergetic beams (beam hardening), scatter, detector blur, smoother filters, iterative reconstruction and motion, none of which are modeled.

Key takeaways

  • Back-projection alone blurs; the ramp filter undoes the blur.
  • Too few angles leave streaks.

Quantum noise and dose

Each ray ends in a photon count \(N\) that is Poisson with mean \(\bar N = N_0 e^{-p}\), where \(N_0\) is the number of photons sent along the ray. To first order,

\[ \operatorname{Var}\!\left[-\ln\frac{N}{N_0}\right] \approx \frac{1}{\bar N} = \frac{e^{p}}{N_0} \tag{7}\]

Rays through more tissue are noisier. Reconstruction is linear, so image noise scales as \(1/\sqrt{N_0}\), and dose is proportional to \(N_0\) (m49): halving the noise costs four times the dose. In lab part D the fitted slope of log noise against log photon count is −0.51.

Key takeaways

  • Noise falls with the square root of dose, not with dose.
  • Thick or dense paths cost the most photons.

ECG gating and motion

A slice needs projections over about 180°. With one X-ray source that takes about half a rotation; with two source-detector pairs 90° apart, about a quarter (Petersilka et al. 2008). For a rotation time \(T_{\text{rot}}\) and \(N_{\text{src}}\) sources, the data window and the blur of a structure moving at speed \(v\) are

\[ \Delta t \approx \frac{T_{\text{rot}}}{2\,N_{\text{src}}}, \qquad b = v\,\Delta t \tag{8}\]

an approximation that ignores the fan angle and the details of how scanners weight their data. It gives 125 ms and 62.5 ms at a 0.25-s rotation, against 125 and 66 ms reported (Mergen et al. 2023), and 82.5 ms at 0.33 s with two sources, against 83 ms (Petersilka et al. 2008). The ECG chooses where in the cycle the window sits.

Coronary arteries moved in plane at 46.6 mm/s on average, from 22.4 mm/s for the left anterior descending artery to 69.5 mm/s for the right coronary artery, and slowest, 27.9 mm/s, at 48% of the cycle (Achenbach et al. 2000). With a 125-ms window that is 5.8 mm of blur at the average speed and 3.5 mm at the slowest phase (lab part E). A disk of diameter \(D\) moving along \(x\) by \(b\) during the window is smeared into

\[ I(x, y) = \frac{\left|\,[x - b/2,\ x + b/2] \cap [-c,\ c]\,\right|}{b}, \qquad c = \sqrt{D^2/4 - y^2} \tag{9}\]

the fraction of the window that the point \((x, y)\) spends inside the disk. Its peak is \(\min(1, D/b)\): a 3-mm vessel at 69.5 mm/s keeps 35% of its brightness with a 125-ms window.

Approximate model

Assumptions

  • Model: Equation 8 and Equation 9, a rigid disk moving at constant in-plane speed throughout the window; tested against epml.ct.
  • Speeds are the published averages for coronary arteries (Achenbach et al. 2000); rotation times of 0.25 and 0.33 s match the cited reports (Mergen et al. 2023; Petersilka et al. 2008), and values in between are illustrative. The vessel diameter is illustrative.
  • What it ignores: speed changes within the window, motion along the scanner axis, the scanner’s own spatial blur, data weighting and any motion correction. It is for intuition, not for predicting image quality.

Key takeaways

  • The window, not the rotation alone, sets the motion blur.
  • A second source or a quieter phase of the cycle shortens the smear.

Thin walls and wall thickness

Model the image blur as a Gaussian with full width at half maximum (FWHM) \(\Gamma\), standard deviation \(\sigma = \Gamma/(2\sqrt{2\ln 2})\). A wall of thickness \(w\) and unit contrast then reads

\[ g(x) = \tfrac12\left[\operatorname{erf}\!\left(\frac{x + w/2}{\sqrt2\,\sigma}\right) - \operatorname{erf}\!\left(\frac{x - w/2}{\sqrt2\,\sigma}\right)\right], \qquad g(0) = \operatorname{erf}\!\left(\frac{w}{2\sqrt2\,\sigma}\right) \tag{10}\]

Thickness read with a fixed threshold at half the contrast vanishes once \(g(0) < 1/2\); read as the width at half of the profile’s own peak, it never drops below \(\Gamma\). With an illustrative 1-mm blur, the thinnest published left atrial wall, 0.5 mm (Beinart et al. 2011), reaches only 44% of its contrast: a half-contrast threshold misses it and a FWHM reading gives 1.06 mm. The mean wall, 1.89 mm, reaches 97% and reads correctly (lab part F).

On a 3-D atrium the wall sits between blood and fat, not between two equal backgrounds, and voxels average whatever they contain (the partial-volume effect). Segment the wall with a window between the fat-wall and wall-blood midpoints. The blurred step from blood to fat passes through that window even where there is no wall at all, over a width set by \(\sigma\) alone: 0.61 mm at 1 mm of blur and 1.21 mm at 2 mm, from the same error function as Equation 10 (lab part G). So with 2 mm of blur the thinnest walls in lab part G (median 0.59 mm) read 1.25 mm, about the width of the step itself. At 1.5-mm voxels the thinnest walls read one voxel, and on some rays behind the atrium the window bridged the 3 mm of fat to the esophagus, so wall and esophagus read as one structure. A distance transform of the segmented atrium gives the distance to the esophagus directly.

Key takeaways

  • A wall thinner than the blur loses contrast before it loses width.
  • Reported thickness depends on blur, voxel size and thresholds as well as on the wall.

Merging CT with the map: two kinds of error

Registration itself is derived in m21; here we only read the evidence. A surface fit reports how far the map’s points lie from the CT surface: 2.3 ± 0.4 mm, against 6.4 ± 2.8 mm for pulmonary-vein landmarks, in 30 patients (Kistler et al. 2006). That number can be small even if the CT is shifted along the surface. In nine dogs with markers sewn to the heart, ablation guided only by the registered CT landed 1.8–2.7 mm from the markers at autopsy, depending on the chamber (Dong et al. 2006). That is the error at a target, the quantity that matters for a lesion. m55 builds an error budget from both.

Key takeaways

  • Ask whether an error is between surfaces or at a target.
  • Validation at a target needs an independent marker, such as one confirmed at autopsy.

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 CT shows up in EP. A European imaging and rhythm consensus document reviews CT alongside echo, MRI and nuclear imaging for patients with AF and highlights its value in planning interventions (Donal et al. 2016). The cited studies use CT to look for appendage thrombus (Romero et al. 2013), to describe the course of the esophagus behind the left atrium (Tsao et al. 2005), to measure atrial wall thickness (Beinart et al. 2011), to study wall thinning around VT isthmuses (Ciaccio et al. 2024; Moore et al. 2024) and to merge anatomy with the map (Kistler et al. 2006; Dong et al. 2006).

What the evidence is. The thrombus figures come from a meta-analysis with TEE as the reference standard (Romero et al. 2013). The anatomy and wall-thickness studies are cross-sectional or small single-center cohorts (Tsao et al. 2005; Beinart et al. 2011; Moore et al. 2024; Ciaccio et al. 2024). The registration figures come from one clinical series and one animal study (Kistler et al. 2006; Dong et al. 2006). None of them is a trial of CT against no CT.

What a reader should ask:

  • Which phase and which window? The atrium changes size and the esophagus moves; a scan is one moment.
  • Which resolution and which thresholds? Thin walls and small distances depend on both (Equation 10).
  • Reported or target error? A surface distance is not the error at a lesion (Kistler et al. 2006; Dong et al. 2006).

Visual summary

Schematic

thorax slice source detector sinogram ramp |k| slice in HU HU: water 0, air −1000 Window ≈ T_rot / (2 N_src) Thin wall: contrast erf(w / 2√2σ) Surface error ≠ target error
Figure 1: CT in one picture: line integrals at many angles form a sinogram, the ramp filter and back-projection rebuild the slice, and the window, the blur and the registration decide what the numbers on it mean.

Assumptions

How experts think

Before trusting a CT number or a measurement made on it, ask:

  1. What energy and what contrast? Hounsfield units depend on the beam and on iodine (see Equation 1 and Equation 2).
  2. Was the motion frozen? Compare the window with the speed of what you are measuring (Equation 8).
  3. Is the structure bigger than the blur? If not, its contrast and its apparent size both change (Equation 10).
  4. How was it measured? Threshold, voxel size and the neighbors of the structure all enter.
  5. Which error is quoted? A surface distance after registration, or the error at a target (Kistler et al. 2006; Dong et al. 2006).

One sentence to keep: a CT image is a reconstruction from line integrals at one moment, at one resolution, on a scale set by water and air.

Cheat sheet

Quantity Result
Hounsfield units (Equation 1) \(1000(\mu - \mu_w)/(\mu_w - \mu_a)\); water 0, air −1000
Iodine (Equation 2) \(\Delta\mathrm{HU} = 1000\,c\cdot10^{-3}(\mu/\rho)_{\mathrm I}/\mu_w\); 36.8 HU per mg/mL at 60 keV
Projection (Equation 3) \(p_\theta(s) = \iint f\,\delta(x\cos\theta + y\sin\theta - s)\)
Central slice (Equation 4) \(P_\theta(k) = F(k\cos\theta, k\sin\theta)\)
Filtered back-projection (Equation 5) filter each projection with \(|k|\), back-project over \([0, \pi)\)
Ram-Lak kernel (Equation 6) \(h[0] = 1/(4d^2)\); \(-1/(\pi^2 n^2 d^2)\) for odd \(n\); 0 for even
Noise (Equation 7) \(\operatorname{Var} \approx e^{p}/N_0\); image noise \(\propto 1/\sqrt{\text{dose}}\)
Gating (Equation 8) \(\Delta t \approx T_{\text{rot}}/(2N_{\text{src}})\); blur \(v\,\Delta t\)
Smeared disk (Equation 9) peak \(\min(1, D/b)\); length \(D + b\)
Thin wall (Equation 10) peak \(\operatorname{erf}(w/(2\sqrt2\sigma))\); \(\sigma = \Gamma/(2\sqrt{2\ln2})\)

Values used in this module

Value Source
Mass attenuation at 60 keV: water 0.2059, iodine 7.579 cm²/g; tissues as in m49 Hubbell and Seltzer (1995)
Window 125 ms (one source) and 66 ms (two) at 0.25 s rotation Mergen et al. (2023)
Window 83 ms with two sources at 0.33 s rotation Petersilka et al. (2008)
Coronary in-plane speed: mean 46.6, 22.4 to 69.5 mm/s by artery, 27.9 mm/s at the slowest phase Achenbach et al. (2000)
Left atrial wall 1.89 mm mean, 0.5–3.5 mm Beinart et al. (2011)

Studies in this module

Study Question Design
Romero et al. (2013) CT for appendage thrombus, TEE as reference Meta-analysis, 19 studies
Tsao et al. (2005) Course of the esophagus behind the left atrium Cross-sectional CT, 48 patients and 32 controls
Beinart et al. (2011) Left atrial wall thickness CT cohort
Moore et al. (2024) Wall thickness at VT isthmuses, repaired tetralogy of Fallot Single-center cohort, 61 patients
Ciaccio et al. (2024) Wall thickness gradients and VT isthmuses after infarction Case series, 6 patients
Kistler et al. (2006) Landmark and surface registration errors Clinical series, 30 patients
Dong et al. (2006) Error at targets after registration Animal study, 9 dogs
Mergen et al. (2023) Image quality at two windows Single-center cohort, 30 patients

Interview questions

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

Q1. What does one CT detector reading measure?

Model answer

The line integral of the attenuation coefficient along the ray, \(-\ln(I/I_0) = \int\mu\,ds\) (m49). All readings at all angles form the Radon transform of the slice (Equation 3).

Q2. Define the Hounsfield unit. Why can the same tissue read differently on two scanners?

Model answer

\(\mathrm{HU} = 1000(\mu - \mu_w)/(\mu_w - \mu_a)\) (Equation 1). Attenuation depends on photon energy, and scanners differ in tube voltage and filtration, so the beam spectrum differs. Reconstruction filters and calibration also matter.

Q3. Why does cardiac CT for mapping need contrast?

Model answer

Blood and myocardium attenuate almost the same: at 60 keV they differ by about 5 HU (lab part A). Iodine adds about 37 HU per mg/mL at that energy (Equation 2), which separates the blood pool from the wall.

Q4. State the central-slice theorem and say why it matters.

Model answer

The 1-D Fourier transform of a projection at angle \(\theta\) equals the 2-D Fourier transform of the image along the line through the origin at that angle (Equation 4). It shows that projections at enough angles determine the image, and it leads directly to filtered back-projection.

Q5. Why does plain back-projection blur, and how does the ramp filter fix it?

Model answer

Projections sample the 2-D spectrum along radial lines, densely near the center and sparsely far out, so plain back-projection over-weights low frequencies (a point becomes a \(1/r\) blur). The polar Jacobian \(|k|\) restores the weighting (Equation 5). In practice the ramp is band-limited (Equation 6) and often smoothed to limit noise.

Q6. You halve the dose. What happens to the image noise?

Model answer

It rises by \(\sqrt2\), about 41%, because noise scales as \(1/\sqrt{N_0}\) and dose is proportional to \(N_0\) (Equation 7).

Q7. What sets temporal resolution in cardiac CT, and how can it be improved?

Model answer

The time to collect about 180° of data: roughly half a rotation with one source, a quarter with two (Equation 8). Faster rotation and a second source shorten it; reported values are 125 and 66 ms at a 0.25-s rotation (Mergen et al. 2023). Choosing a quiet phase of the cycle lowers the speed instead (Achenbach et al. 2000).

Q8. A vessel looks elongated and faint on a gated scan. Explain with numbers.

Model answer

Motion during the window. At 69.5 mm/s and 125 ms it moves 8.7 mm, so a 3-mm vessel is smeared to about 11.7 mm and keeps \(3/8.7 \approx\) 35% of its peak brightness (Equation 9).

Q9. Two groups report different mean left atrial wall thickness from CT. Give engineering reasons.

Model answer

Different blur and voxel size, different thresholds or reading methods (fixed threshold against FWHM), different cardiac phase, and different places on the wall. A wall near the blur width loses contrast and either vanishes or reads too thick depending on the method (Equation 10); published walls are 0.5–3.5 mm (Beinart et al. 2011).

Q10. A mapping system reports a registration error of 2 mm. Does that mean lesions land within 2 mm of where the CT says?

Model answer

No. The reported number is usually the distance between map points and the CT surface after the fit (Kistler et al. 2006); a shift along the surface does not change it. The error at a target must be measured with an independent reference, as in the animal study that found 1.8–2.7 mm by chamber (Dong et al. 2006).

Q11. How would you measure the distance from the posterior left atrium to the esophagus on CT?

Model answer

Segment both, compute a Euclidean distance transform of the atrial segmentation and take its minimum over the esophagus voxels (lab part G). State the phase, the voxel size and the thresholds, because partial volume can merge the two when the fat between them is thin.

Q12. CT found thrombus with 92% specificity. Why were most positive scans false positives in the meta-analysis?

Model answer

Because thrombus was uncommon: at 8.9% prevalence, 96% sensitivity and 92% specificity give a positive predictive value near 41% (Romero et al. 2013). Predictive values depend on prevalence (m35); a delayed scan raised specificity to 99% and the positive predictive value to 92%.

Code lab

The lab goes from tabulated attenuation to a reconstructed slice and ends with wall thickness on a 3-D atrium.

Part What you do Data
A Hounsfield units and iodine (Equation 1, Equation 2) NIST attenuation tables
B Projections and the central-slice theorem (Equation 3, Equation 4) our own phantom
C Filtered back-projection (Equation 6, Equation 5) our own phantom
D Quantum noise and dose (Equation 7) our own phantom
E The gating window and motion blur (Equation 8, Equation 9) published scanner and motion values
F Thin walls under blur (Equation 10) published wall thickness range
G Wall thickness and esophagus distance on a 3-D left atrium our own phantom

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 voxel has \(\mu = 1.1\,\mu_w\). What is its CT number?

Answer

\(1000 \times 0.1 = 100\) HU (Equation 1, with \(\mu_a = 0\)).

S2. At 37 HU per mg/mL, what iodine concentration lifts blood by 300 HU?

Answer

About 8 mg/mL (Equation 2).

S3. Every projection of an image has the same integral over \(s\). Why?

Answer

Integrating \(p_\theta(s)\) over \(s\) integrates \(f\) over the whole plane, whatever the angle (Equation 3). It is also the \(k = 0\) value in Equation 4.

S4. A single-source scanner rotates in 0.30 s. About how long is its half-scan window?

Answer

About 150 ms (Equation 8).

S5. You need half the noise. How much more dose, all else equal?

Answer

Four times (Equation 7).

S6. True or false: a wall thinner than the blur always looks thinner than it is.

Answer

False. With a fixed threshold it may vanish; read as a width at half its own peak it looks at least as wide as the blur, so thicker than it is (Equation 10).

Literature figures

These sources are worth looking at alongside this module. We link to them instead of copying their figures: we could not confirm a license that allows reuse here. Each entry says what to look for and where this module explains it.

  • Filtered back-projection, from the source. Kak and Slaney’s book is free to read for personal use on the authors’ site; see its chapter on reconstruction algorithms for nondiffracting sources and its chapter on aliasing artifacts and noise in CT images [Kak and Slaney (2001); electronic copy]. This module: filtered back-projection.
  • Esophageal routes. Tsao and colleagues classify the course of the esophagus by its nearest pulmonary vein and report the distances [Tsao et al. (2005); article]. This module: from the field.
  • Error at the target. Dong and colleagues compare ablation sites with fiducial markers, chamber by chamber [Dong et al. (2006); article]. This module: merging CT with the map.

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.
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.
Ciaccio, Edward J., Nicolas Cedilnik, Henry H. Hsia, et al. 2024. “Wavefront Curvature Analysis Derived from Preprocedural Imaging Can Identify the Critical Isthmus in Patients with Postinfarcted Ventricular Tachycardia.” Heart Rhythm 21 (12): 2471–80. https://doi.org/10.1016/j.hrthm.2024.05.056.
Donal, Erwan, Gregory Y. H. Lip, Maurizio Galderisi, et al. 2016. “EACVI/EHRA Expert Consensus Document on the Role of Multi-Modality Imaging for the Evaluation of Patients with Atrial Fibrillation.” European Heart Journal: Cardiovascular Imaging 17 (4): 355–83. https://doi.org/10.1093/ehjci/jev354.
Dong, Jun, Hugh Calkins, Stephen B. Solomon, et al. 2006. “Integrated Electroanatomic Mapping with Three-Dimensional Computed Tomographic Images for Real-Time Guided Ablations.” Circulation 113 (2): 186–94. https://doi.org/10.1161/CIRCULATIONAHA.105.565200.
Hubbell, J. H., and S. M. Seltzer. 1995. Tables of x-Ray Mass Attenuation Coefficients and Mass Energy-Absorption Coefficients 1 keV to 20 MeV for Elements z = 1 to 92 and 48 Additional Substances of Dosimetric Interest (NIST Standard Reference Database 126). National Institute of Standards and Technology. https://physics.nist.gov/PhysRefData/XrayMassCoef/.
Kak, Avinash C., and Malcolm Slaney. 2001. Principles of Computerized Tomographic Imaging. Vol. 33. Classics in Applied Mathematics. Society for Industrial; Applied Mathematics. https://doi.org/10.1137/1.9780898719277.
Kistler, Peter M., Mark J. Earley, Stuart Harris, et al. 2006. “Validation of Three-Dimensional Cardiac Image Integration: Use of Integrated CT Image into Electroanatomic Mapping System to Perform Catheter Ablation of Atrial Fibrillation.” Journal of Cardiovascular Electrophysiology 17 (4): 341–48. https://doi.org/10.1111/j.1540-8167.2006.00371.x.
Mergen, Victor, Thomas Sartoretti, Giulia Cundari, et al. 2023. “The Importance of Temporal Resolution for Ultra-High-Resolution Coronary Angiography: Evidence from Photon-Counting Detector CT.” Investigative Radiology 58 (11): 767–74. https://doi.org/10.1097/RLI.0000000000000987.
Moore, Jeremy P., Jonathan Su, Kevin M. Shannon, et al. 2024. “Multidetector Computed Tomography Assessment of Anatomical Ventricular Tachycardia Isthmuses in Repaired Tetralogy of Fallot.” JACC: Clinical Electrophysiology 10 (5): 857–66. https://doi.org/10.1016/j.jacep.2024.102333.
Petersilka, Martin, Herbert Bruder, Bernhard Krauss, et al. 2008. “Technical Principles of Dual Source CT.” European Journal of Radiology 68 (3): 362–68. https://doi.org/10.1016/j.ejrad.2008.08.013.
Radon, J. 1986. “On the Determination of Functions from Their Integral Values Along Certain Manifolds.” IEEE Transactions on Medical Imaging 5 (4): 170–76. https://doi.org/10.1109/TMI.1986.4307775.
Romero, Jorge, Syed Arman Husain, Iosif Kelesidis, et al. 2013. “Detection of Left Atrial Appendage Thrombus by Cardiac Computed Tomography in Patients with Atrial Fibrillation: A Meta-Analysis.” Circulation: Cardiovascular Imaging 6 (2): 185–94. https://doi.org/10.1161/CIRCIMAGING.112.000153.
Tsao, Hsuan-Ming, Mei-Han Wu, Satoshi Higa, et al. 2005. “Anatomic Relationship of the Esophagus and Left Atrium: Implication for Catheter Ablation of Atrial Fibrillation.” Chest 128 (4): 2581–87. https://doi.org/10.1378/chest.128.4.2581.