Recognising and salvaging artifacts

The core idea

Optimisation means diagnostic information—not cosmetic perfection

The best CTCA is not necessarily the prettiest. It is the examination that answers the clinical question with the least reasonable radiation, contrast, delay, and repeat exposure. Salvage begins by proving what kind of artifact is present.

Dr Pandey’s practical sequence is: classify the motion source → name the cardiac artifact → inspect the ECG and acquired window → apply the least invasive rescue → verify the result in source images and multiple planes.

Before touching reconstruction controls

Confirm the problem and its extent

1. Check identity and protocol: coverage, scan mode, ECG window, phase, slice thickness, kernel, contrast enhancement, and dose record.
2. Scroll the thin axial series: locate every affected z-axis level and decide whether the defect is continuous, focal, periodic, or phase-dependent.
3. Review multiplanar views: especially a sagittal reconstruction including chest wall, ribs, diaphragm, spine, aortic root, and heart.
4. Open the acquisition ECG: look for rate drift, ectopy, pauses, noise, missed R waves, tall T waves, and the scanner’s synchronization markers.
5. Map clinical consequence: which coronary segments or other targets are truly non-diagnostic? Do not repeat an examination for an irrelevant cosmetic defect.

The sagittal test

Chest wall and spine reveal the source of motion

Diaphragmatic duplication and periodic bands through the lungs support breathing. A break confined to the aortic root or coronary artery with stable skeletal landmarks supports cardiac origin. Use the whole volume; a single coronary reformat can hide the diagnostic clue.

Cardiac pattern 1

Blur, ghosting, streaking, or “winging”

A rapidly moving vessel is averaged over the temporal reconstruction window, producing an indistinct wall, a comma-shaped smear, double edge, or trailing “wing.” The vertical mid-RCA is particularly vulnerable because it moves rapidly, especially during phases of atrial contraction and filling.

• Clue: local loss of edge sharpness without a discrete slab boundary; the adjacent stationary chest remains sharp.
• Mimic: can simulate noncalcified plaque or make the lumen appear narrowed.
• First rescue: reconstruct another available phase—often systole when diastole is blurred, or an adjacent diastolic phase for a single vessel.
• Other options: vendor-validated motion compensation, a different temporal reconstruction if supported, and careful orthogonal confirmation.

If every available phase is blurred, post-processing cannot create temporal resolution that was not acquired. Mark the affected segment non-diagnostic rather than “smoothing” it into apparent normality.

Cardiac pattern 2

Transition, stair-step, or misregistration artifact

In sequential/step-and-shoot acquisition, the heart may be covered in several slabs acquired over different beats. If heart rate or phase differs between adjacent slabs, anatomy shifts abruptly at the join. Vertically oriented structures show a step; a horizontal coronary crossing the transition zone can look focally pinched and mimic severe stenosis.

• Clue: a straight or curved boundary across the image with abrupt offset of several structures at the same z-axis level.
• High-risk mimic: an apparent focal LAD stenosis exactly at a slab junction.
• Rescue principle: the transition zone itself may persist, but the moving coronary occupies a different position in another cardiac phase. Combine diagnostic information from complementary phases.
• Software: vendor-specific misregistration correction may reduce the step but must be validated against thin source images.

Cardiac pattern 3

Interpolation artifact: missing information is smeared across a gap

Retrospectively gated helical reconstruction can face a data gap when the patient’s heart rate slows relative to pitch, after a premature beat and compensatory pause, or when a true R wave is missed. The algorithm interpolates between available data, producing smearing, stretched anatomy, or an apparent absent band.

• Clue: broad smear or discontinuity that may move to different z-axis levels across reconstructed phases.
• ECG clue: a long R–R interval, pause, small missed R wave, or missing synchronization marker.
• Likely rescue: if a true R wave was missed, add/move the synchronization point correctly; if the rhythm itself created a long gap, experienced vendor-specific ECG composing may regularise reconstruction.
• Limitation: do not assume interpolation has created “real” tissue—verify the reconstructed anatomy in adjacent phases and planes.

Cardiac pattern 4

Duplication artifact: the same anatomy is represented twice

When heart rate is fast relative to pitch, after a premature beat, or when the scanner mistakes a tall T wave/noise spike for an R wave, overlapping cardiac information may be assigned to separate cycles. A valve, aortic root, coronary segment, or chamber edge can appear doubled.

• Clue: two copies or parallel edges of the same structure rather than simple blur.
• ECG clue: an extra synchronization point, short R–R interval, abrupt tachycardia, or ectopic beat.
• Likely rescue: remove a false marker placed on a T wave/noise spike; for true irregularity, vendor-specific composing may exclude or redistribute the abnormal beat.
• Danger: duplicated lumen edges can resemble dissection, valve pathology, an anomalous branch, or severe stenosis.

Read what the scanner believed

The synchronization point is a machine label—not proof of an R wave

The scanner places a synchronization marker at what it interprets as the start of each cardiac cycle. The goal for ordinary ECG editing is one marker on every true R wave and nowhere else. Errors include a small R wave being missed, a tall T wave or noise spike being labelled as R, a bifid R receiving two markers, or a marker being shifted away from the QRS.

Always preserve the original reconstruction and ECG record. Create a separately labelled edited series and compare it with unedited data; do not erase evidence of the original acquisition.

Rescue tool 1

Multiphase reconstruction: search the acquired window systematically

Retrospective ECG gating commonly provides broad phase coverage; prospective scans provide only the exposed phase window plus any padding. Reconstructing at small phase increments can find a motion-free interval for one or more coronary segments.

1. Determine which phases actually received adequate tube current; low-current phases may be available but excessively noisy.
2. Start near expected quiet diastole and end-systole, then sample nearby intervals in small steps appropriate to the scanner.
3. Compare the same coronary segment side by side across phases—not only the whole-heart appearance.
4. Use complementary phases when a transition boundary affects different vessel levels.
5. Save the diagnostic phase(s) with clear labels and document residual limitations.

Multiphase rescue cannot access a phase outside a narrow prospective exposure window. It also does not repair respiratory/body motion, absent contrast, or omitted anatomy.

Rescue tool 2

Absolute-millisecond systolic reconstruction stabilises variable R–R timing

A percentage is relative to cycle length: 70% occurs at 700 ms when R–R is 1000 ms, but at 420 ms when R–R is 600 ms. Because diastole shortens disproportionately as heart rate rises, the same percentage can land in different mechanical phases from beat to beat.

Reconstructing at a fixed interval after the R wave—often exploring end-systole around 250–350 ms—can align the heart more consistently during tachycardia, atrial fibrillation, or marked rate variability. The exact millisecond value is patient- and scanner-specific; inspect several candidates rather than treating 300 ms as universal.

Rescue tool 3

ECG sync-point editing corrects detection errors

Editing aligns scanner markers with the ECG that was actually recorded. It is appropriate when the rhythm may be regular but the scanner misidentified it.

• Add: a marker on a genuine missed R wave to recover a correctly delimited cycle and reduce interpolation.
• Delete: a false marker on a T wave, bifid component, or noise spike to remove an artificial cycle and reduce duplication.
• Move: a misplaced marker onto the true R wave.

After editing, repeat reconstruction at the selected phase and compare before/after images at exactly the same anatomy. If the corrected series creates new breaks or changes anatomy implausibly, step back. Interface names and permitted edits differ by vendor; use trained local workflow.

Rescue tool 4

ECG composing deliberately changes cycle boundaries

In true arrhythmia, the scanner may have marked every R wave correctly yet the intervals are unsuitable for a consistent multi-beat reconstruction. “Composing” synthetically redistributes boundaries—for example, removing a marker associated with a very short premature beat or adding an artificial marker within a long post-ectopic pause—to make reconstruction intervals more regular.

This is fundamentally different from correcting a detection error: a composed marker may not represent a true R wave, and a true R-wave marker may be intentionally excluded. The aim is to exploit redundant helical data and reduce a severe interpolation or duplication defect.

Not every defect is motion

Recognise other common limitations and use targeted salvage

Every rescue has a trade-off. Thicker/smoother data reduce noise but also spatial resolution; sharper kernels reduce blooming but increase noise; low-energy VMI raises iodine conspicuity but can amplify noise.

A disciplined order of operations

The CTCA salvage ladder

1. Localise: identify exactly which anatomy and z-axis levels are degraded.
2. Classify: body, respiratory, cardiac, contrast, noise, beam hardening, blooming, or post-processing.
3. Inspect evidence: sagittal/coronal images, thin axial stack, acquisition ECG, heart-rate trace, bolus information, and protocol.
4. Try reversible display changes: window/level, plane, centreline, slab thickness.
5. Reconstruct from unchanged ECG: alternate phase, absolute milliseconds, thickness, kernel, iterative/spectral option.
6. Correct genuine ECG detection errors: add, remove, or move markers on true/false R-wave labels.
7. Consider advanced composing: only for true irregularity with suitable helical data and expert validation.
8. Compare: original versus rescued series at identical anatomy; look for newly introduced artifacts.
9. Decide: diagnostic, diagnostic with limitation, segment non-diagnostic, or repeat acquisition clinically justified.

The final safety decision

Repeat only when salvage fails and the result will change management

Gross body or respiratory motion often cannot be corrected after acquisition. Even so, a full repeat is not automatic. Determine whether the clinically relevant vessels are diagnostic, whether an alternative test is preferable, and whether a carefully planned limited repeat can answer the missing question without duplicating the whole examination.

• Before repeating: identify the cause, correct breath-hold/positioning/ECG/heart-rate or IV issue, recalculate contrast and timing, and confirm the patient can cooperate.
• Balance harm: additional radiation, contrast, medication, time, and the risk of another failed scan versus the clinical value.
• Document: the non-diagnostic segments, reason, salvage attempted, residual confidence, and repeat/alternative recommendation.

A reconstruction that looks better but is not anatomically trustworthy is not a successful salvage.

Close the loop

Use every failed pattern to improve the next acquisition

Review recurring artifacts by scanner and protocol. If transition bands repeatedly cross the same coronary level, or T-wave oversensing recurs with a lead configuration, the best solution is a protocol/ECG setup change—not repeated post-processing heroics.

Apply the algorithm

Worked artifact scenarios

Case A · Broken coronary plus displaced spine and chest wall

This is gross body motion, not a cardiac gating problem. ECG editing is unlikely to help. Determine whether the relevant anatomy remains diagnostic; if not, address the cause and weigh a repeat or alternative test.

Case B · Periodic diaphragm/chest-wall bands with stable spine

The pattern supports respiratory motion. Do not waste time composing the ECG. Check whether unaffected phases/levels answer the question; otherwise repeat only after better breath-hold coaching and a shorter suitable acquisition if available.

Case C · Mid-RCA winging at 75%; skeleton stable

Classify as cardiac blur. Reconstruct nearby available phases, including an end-systolic series when appropriate, and compare the exact RCA segment. Motion compensation may be an adjunct, not independent proof.

Case D · Apparent proximal LAD stenosis exactly at a slab boundary

Suspect transition misregistration. Inspect other structures at the same z-axis level and reconstruct a complementary phase in which the LAD has moved away from the transition. Confirm in thin axial and orthogonal images.

Case E · Doubled aortic root; ECG markers occur on both QRS and tall T waves

This is duplication from false synchronization points. Remove only the markers on T waves using the validated interface, reconstruct a separate series, and verify normal anatomy against the original and multiple planes.

Case F · AF with correct but irregularly spaced R-wave markers

This is not a marker-detection error. Explore absolute-millisecond systolic reconstruction first. If severe interpolation/duplication persists and suitable retrospective data exist, expert ECG composing may be considered and rigorously validated.

Active recall

Self-test: answer before opening each explanation

Printable quick revision

One-page artifact salvage card

• Blur: smeared/winged vessel; search another available phase.
• Transition: slab-junction offset; use complementary phases and check same-level structures.
• Interpolation: missing data smeared; look for a pause or missed R-wave marker.
• Duplication: anatomy doubled; look for extra marker, short beat, or abrupt tachycardia.
• % R–R: same percentage is a different clock time when rate changes.
• Absolute ms: fixed systolic time can stabilise variable cycles—but requires correct R-wave labels.
• Editing: correct marker errors. Composing: deliberately alter cycle boundaries in true irregularity.
• Non-motion: tailor thickness/kernel/iterative/VMI/windowing to noise, contrast, beam hardening, or blooming.
• Final rule: compare original and salvaged series; never hide residual uncertainty.

At the workstation, ask: What moved, what data were actually acquired, and could this “fix” create a false anatomy?

Provenance

Sources and scope

This chapter follows Dr Niraj Pandey’s Optimizing outcomes in coronary CT angiography Webinar 2 talk, especially approximately 36:15–70:46. Automated captions contain transcription errors; terminology and rescue principles were checked against current professional guidance.

1. SCCT Interpretation and Reporting of Coronary CTA: 2026 Update — current artifact classification, alternate phases, absolute-ms reconstruction, ECG editing, contrast/noise salvage, and interpretation limits.
2. Practical Tips and Tricks for Salvaging ECG-gated Cardiac CT from Motion Artifacts — body/respiratory/cardiac pattern recognition and reconstruction rescue.
3. Artifacts at Cardiac CT: Physics and Solutions — motion, partial volume, beam hardening, metal, noise, and ECG-editing principles.
4. SCCT Guidelines for the Performance and Acquisition of Coronary CTA (2016) — acquisition windows, padding, phase reconstruction, and dose trade-offs.

Educational material only. ECG-editing/composing interfaces and allowable reconstructions vary by vendor and acquisition mode. Preserve original data, use trained local workflow, and report non-diagnostic segments honestly. External references require internet; this chapter and self-test work offline.