Artifacts by Modality
- An artifact is anything in the image that isn't actually in the patient — the machine's fingerprints, not the disease.
- Most artifacts are predictable from how each modality makes its picture, so the same physics that creates the image creates its lies.
- The danger isn't ugliness; it's an artifact that mimics pathology (or hides it). Knowing the usual suspects keeps you from calling a glitch a tumor.
- A quick sanity check: does this "finding" obey anatomy, or does it follow the equipment? Artifacts tend to line up with the machine, not the body.
Every imaging machine is a storyteller, and like any storyteller it occasionally embellishes. An artifact is the embellishment: a feature on the image that doesn't correspond to anything real in the patient. The trap is that some artifacts are gorgeous liars — they look exactly like the thing you're scared of missing. The good news is that artifacts aren't random. Each modality cheats in its own characteristic way, so once you know how a machine builds its image, you can usually predict the specific fibs it will tell.
Think of it like recognizing whose handwriting you're looking at. CT, MRI, and ultrasound each have their own scrawl. Let's learn the penmanship.
CT: the modality that hates metal
CT builds its picture by shooting X-rays through you from every angle and doing math to figure out how dense each spot is. That math leans hard on the assumption that the beam behaves itself. Metal does not let it behave itself.
A hip prosthesis, dental fillings, or surgical clips absorb so much of the beam that the detectors on the far side basically get nothing, and the reconstruction — read more about how that math works in CT physics & reconstruction — panics. The result is streak artifact: bright and dark rays fanning out from the metal like a cartoon sunburst, smearing anatomy right next to the hardware. That's also exactly where you'd want to look for, say, loosening or infection. Inconvenient.
CT's other classic is beam hardening, where the beam loses its softer X-rays as it passes through dense bone and comes out "harder." Between two dense structures — the petrous bones at the skull base are the famous offenders — you get dark bands. A dark streak across the brainstem can do a remarkable impression of a stroke or hemorrhage if you take it at face value.
A dark band crossing the brainstem on a head CT is usually beam hardening, not infarct or bleed. The tell: it runs in a straight line between two chunks of dense bone and ignores the actual shape of the brainstem. Real pathology respects anatomy; this artifact respects the bones.
MRI: a modality with a lot of feelings
MRI makes its image by listening to spinning protons, which means it is exquisitely sensitive to anything that disturbs the magnetic field or the timing. The cost of all that sensitivity is a long list of artifacts. (If the T1/T2 vocabulary is hazy, a quick detour through MRI basics helps.)
The one everyone meets first is motion artifact. Because MRI builds the image over many seconds, a patient who breathes, swallows, or fidgets smears the data — producing ghostly repeated copies that march across the image in one direction (the "phase-encode" direction). A pulsating vessel does the same thing, flinging faint duplicate aortas across the abdomen.
Then there's chemical shift artifact, which shows up at the borders between fat and water (think kidney against perinephric fat). Fat and water protons resonate at slightly different frequencies, so the scanner misplaces the fat signal by a pixel or two, drawing a crisp black or white line along that interface that can fake a real edge. And susceptibility artifact — metal again, or air, or blood products — distorts the local field and blooms into a dark void, which is genuinely useful for spotting tiny hemorrhages but can swallow nearby anatomy whole.
| Artifact | Born from | Looks like | Usual scene of the crime |
|---|---|---|---|
| Streak | Metal blocking the beam | Bright/dark sunburst rays | CT near hardware, fillings |
| Beam hardening | Beam "hardening" through bone | Dark bands between dense bone | CT at skull base, posterior fossa |
| Motion / ghosting | Patient or vessel moving during scan | Repeated faint copies in one direction | MRI, especially abdomen and chest |
| Chemical shift | Fat vs water resonating differently | Crisp black/white line at fat-water borders | MRI around kidneys, orbits |
| Susceptibility | Metal/air/blood warping the field | Dark blooming void with distortion | MRI near implants, blood products |
If you want the wider toolkit of MRI quirks, the MRI sequences page covers how acquisition choices invite or tame each one.
Ultrasound: shadows, mirrors, and tails
Ultrasound paints with sound, so its artifacts are all about how sound bounces, gets absorbed, or gets confused — the physics lives in ultrasound physics. A gallstone or bone reflects almost everything, leaving a dark acoustic shadow behind it; a fluid-filled cyst lets sound sail through easily, leaving a bright stripe of posterior acoustic enhancement deep to it. Both are artifacts — and both are wonderfully helpful, because they tell you what a structure is made of.
Less helpful are mirror-image artifacts (a structure duplicated across a strong reflector like the diaphragm, conjuring a "lesion" above it that isn't there) and reverberation, where sound ping-pongs between two close reflectors and draws a ladder of false echoes.
Ultrasound artifacts are a big enough topic to deserve their own page — for the full menagerie of shadows, comet tails, and ring-down, see ultrasound artifacts.
How to not get fooled
The unifying trick across every modality is a single question: does this finding follow the patient, or the machine? A true lesion sits where anatomy says it should, persists across views, and behaves consistently. An artifact lines up with the hardware — the metal, the bone, the diaphragm, the scan direction — and often vanishes or shifts when you change the projection, the sequence, or the probe angle.
When something looks weird, ask what the machine was doing there, not just what the disease might be. Half of "interesting findings" are just the scanner showing its work.
Artifacts also intertwine with plain image quality — noise, resolution, and contrast all change how loudly an artifact shouts; that's covered in resolution, noise & contrast. Learn the handful of usual suspects above, keep asking whether a "finding" obeys anatomy or equipment, and you'll spend a lot less time chasing ghosts the machine invented.