Reading CBCT artifacts: the 5 most common patterns and what causes them

Every CBCT image you read contains artifacts. Some are innocuous; some can mimic pathology and derail your diagnosis. Being able to recognize the five most common patterns separates a confident reader from a cautious one.

This is a field guide. For each artifact: what it looks like, what causes it, when to re-acquire and when to read around it.

1. Beam hardening (dark streaks from dense material)

What you see: dark bands or streaks radiating from dense objects. Most common source in dentistry: metal restorations, crowns, pins, implants themselves, orthodontic appliances.

What causes it: the X-ray beam is polyenergetic. Lower-energy photons are absorbed more readily by dense material, so the beam "hardens" (gets biased toward higher energies) as it passes through. The reconstruction algorithm assumes a uniform beam, so it underestimates attenuation in the shadow of the metal.

Classic mimic: the dark streak can look like bone loss or a root fracture near a crown or post.

How to read around it: look at the orthogonal plane (if the streak is visible axially, check coronal and sagittal). True pathology is present in multiple planes; a streak is directional. Most modern scanners offer a metal artifact reduction (MAR) algorithm — run it and compare.

Re-acquire?: only if the streaks obscure the specific diagnostic question. For an implant case far from the affected teeth, often no. For a root fracture evaluation adjacent to a crown, sometimes yes, with a smaller FOV.

2. Motion artifact (blurred or doubled cortical edges)

What you see: cortical outlines look fuzzy or duplicated. Worst affected region is typically the mandible because it's the most movable structure during the 8–25 second acquisition.

What causes it: patient moved during the scan. Could be a cough, a swallow, a head shift, or an unstable positioning.

Classic mimic: mild motion can look like a cortical breach or subtle fracture. Severe motion makes the entire scan non-diagnostic.

How to read around it: scroll through adjacent axial slices. True cortical breach is localized; motion affects many slices consistently in the same direction. 3D rendering exaggerates motion — check the render for "double vision" of the mandibular border.

Re-acquire?: yes if the study's diagnostic question depends on cortical detail. Prevention is cheaper: use a chin rest, head strap and bite block. Remind the patient to stay still and to breathe gently, not to swallow.

3. Scatter and cupping (low-contrast haze, darker centers)

What you see: a "cupping" effect where the center of a structure looks darker than the periphery, or a low-contrast haze across the volume. Often more pronounced in large-FOV acquisitions.

What causes it: X-rays scatter within the patient's tissue and within the scanner itself. The detector reads a mix of direct and scattered photons, and the reconstruction can't fully separate them. Larger FOVs expose more tissue to the beam and generate more scatter.

Classic mimic: cupping can look like a diffuse density change. Rarely mimics specific pathology, but reduces overall image quality.

How to read around it: scroll window/level settings actively. A real density change will respond to windowing in a physiologically plausible way; scatter responds but the gradient looks unnatural.

Re-acquire?: rarely. Modern scanners apply scatter correction automatically. If scatter is severe, the scan was poorly calibrated and should go to service.

4. Ring artifact (concentric circles)

What you see: perfectly concentric rings in the axial plane, often best seen in air around the patient but also visible through tissue.

What causes it: a detector pixel is miscalibrated or dead. The pixel appears in every projection at the same geometric offset from the rotation center, which reconstructs to a ring.

Classic mimic: rarely mimics pathology directly but can obscure small findings along the ring's path.

How to read around it: note the ring's position and mentally exclude it from interpretation along that circle. Findings exactly on the ring's path should be confirmed in a second plane.

Re-acquire?: no. Fix the scanner. Run the manufacturer's detector calibration routine (usually a 5-minute procedure). Persistent rings after calibration mean a service visit is needed.

5. Aliasing and stair-step (jagged edges in oblique or 3D views)

What you see: jagged, stair-stepped edges on curved surfaces when viewed in oblique reformats or 3D renders. Worse at low voxel resolution.

What causes it: voxel size is too coarse for the structure you're viewing. When the display has to interpolate between widely spaced voxels, edges that should be smooth appear stepped.

Classic mimic: can look like a rough cortical surface or irregular implant thread.

How to read around it: check voxel size in the DICOM header. If it's 0.4–0.5 mm, stair-step is expected in oblique views. Measure in the native axial plane, not the reformatted oblique, for highest accuracy.

Re-acquire?: only if the diagnostic question requires sub-mm precision (some endo and surgical cases). Increase the protocol to 0.1–0.2 mm voxel for those specific indications.

A 30-second artifact check routine

Every CBCT, before interpretation:

1. Scroll axial top to bottom. Any metal streaks? Any motion blur?
2. Check 3D render. Rings? Double cortical outlines?
3. Note voxel size and FOV. Does it match the clinical question?
4. Decide: read as-is, read around an artifact, or re-acquire?

When in doubt, correlate

An ambiguous finding on a CBCT is not a diagnosis. Correlate with:

• Clinical examination (tender to percussion? mobile? pus?)
• A second imaging modality (periapical radiograph, photograph, intraoral scan)
• A repeat CBCT at a different FOV or voxel setting if the question is specific

Summary

Five artifacts account for most of what you'll see on any CBCT: beam hardening, motion, scatter, ring, and aliasing. Each has a signature look and a different fix. Prevention at acquisition saves diagnostic headaches later.

Build a quick check into every read. Thirty seconds up front catches the scans that would otherwise lead you to the wrong answer.