Imaging Nerd

Quality Control & Image Artifacts

Key Points
  • In nuclear medicine, the "camera" sees one photon at a time, so almost every artifact is a story about a photon that got lost, blocked, miscounted, or sent to the wrong address.
  • Many of the scariest-looking artifacts are mechanical: a cracked crystal, a dead detector element, or a patient who moved — not disease.
  • Attenuation (photons eaten on the way out) is the great impersonator: it can fake a perfusion defect, especially in the heart, breasts, and diaphragm.
  • A good tech does daily quality control (uniformity, resolution) precisely so the artifact gets caught on a test pattern instead of on your patient.
  • When something looks weird, the first question isn't "what disease is this?" — it's "is this even real?"

Most of imaging is about catching light. Nuclear medicine is the oddball cousin that catches gamma rays — individual little packets of energy fired off by a tracer you injected into the patient. The gamma camera sits outside the body counting those photons one by one and slowly builds a picture out of the tally, like making a pointillist painting where every dot is a single radioactive sneeze.

The catch: a lot can happen to a photon between the patient and the detector. And when photons go missing or land in the wrong spot, you don't get noise that looks like noise — you get a clean, convincing dark spot that looks exactly like real disease. That's the whole pitfall of this page. The artifacts here don't look like artifacts. They look like answers.

The big one: attenuation

The single most important fake-out in nuclear medicine is attenuation — fancy word for "photons getting absorbed or scattered before they ever reach the camera." A gamma ray born deep in the chest has to claw its way out through muscle, bone, and fat, and plenty don't make it. The camera can't count what never arrives, so that region simply looks cold (low counts).

Now here's the trap: a true defect — say, scar in the heart muscle — also looks cold. The camera can't tell the difference between "no tracer here" and "lots of tracer here, but the signal got eaten on the way out."

Pitfall

The classic offenders in myocardial perfusion imaging: breast tissue creating an anterior wall "defect," and the diaphragm/abdomen attenuating the inferior wall. Both can mimic ischemia in a perfectly healthy heart. The fix is attenuation correction (often a CT map on SPECT/CT) and checking that the "defect" is fixed, matches an anatomic shadow, and doesn't track with a real coronary territory.

The tell is that an attenuation artifact stays put — it looks the same on stress and rest, because anatomy doesn't change between the two scans, while ischemia does.

Figure · SPECT
Stress and rest myocardial perfusion SPECT short-axis slices showing a fixed inferior wall count reduction unchanged between stress and rest, consistent with diaphragmatic attenuation rather than a true fixed defect; include the attenuation-corrected series showing normalization of the inferior wall.

Artifacts from the machine itself

Before you blame the patient's physiology, blame the hardware. The detector is a large flat crystal (usually sodium iodide) that flashes when a gamma ray hits it. If something is wrong with the crystal or the electronics, every image inherits the flaw.

  • Non-uniformity. Ideally, flood the camera with an even source and you get an even gray field. In reality, dead or drifting detector regions show up as focal hot or cold spots in that field. On SPECT, a single off region gets smeared by the rotation into a ring artifact — a perfect bullseye that no disease draws.
  • Cracked crystal. Sodium iodide is hygroscopic and brittle; a thermal shock or a bump can crack it. The result is a stark, geometric photopenic line or wedge — sharp edges that biology never makes.
  • Center-of-rotation error. If the camera's assumed center of spin drifts from the true one, SPECT reconstruction blurs points into little donuts or tuning-fork shapes.
Note

This is exactly why techs run daily quality control: a uniformity flood, plus periodic resolution and center-of-rotation checks. The whole point is to catch the cracked crystal on a test pattern at 7am, not on someone's bone scan at noon.

Artifacts from the patient and the tracer

The patient is not a cooperative photon source. They breathe, they shift, they have plumbing.

ArtifactWhat you seeWhat's really going on
Patient motionBlurring or split/doubled structures (esp. SPECT)Patient moved mid-acquisition
Free pertechnetateUnexpected stomach, thyroid, salivary uptakeTracer broke down; free Tc-99m wandered off
Injection infiltrationHot blob at the injection site + faint everything elseDose leaked into soft tissue instead of vein, so less reached the body
Urine/sweat contaminationA random hot spot on the skin or sheetsSpilled radioactive urine, not a lesion

That infiltration one is sneaky: a partially infiltrated dose means fewer counts reach the target organ, so a bone scan can look falsely faint and underwhelming everywhere. Always glance at the injection site before calling a study "low uptake."

Clinical Pearl

Free technetium has a signature distribution: thyroid, salivary glands, and stomach all light up because those tissues naturally concentrate the pertechnetate ion. If you see that trio where it doesn't belong, suspect a tracer problem — the radiopharmaceutical came apart — not a flurry of new pathology.

The one habit that saves you

Every other modality lets you ask "what disease is this?" first. Nuclear medicine makes you earn that question. Because the image is built from counted photons, the first reflex on anything odd should be: is this real, or did a photon get lost? Check stress-versus-rest, check the injection site, check whether the "lesion" sits on a known attenuator or repeats with the camera's geometry.

Get that reflex right and the rest of nuclear medicine gets a lot less spooky — most of the monsters under the bed turn out to be a tired crystal and a patient who couldn't hold still.