MRI Artifacts (Chemical Shift, Aliasing, Motion, Susceptibility)
- An MRI artifact is the scanner confidently drawing something that isn't really there — your job is to recognize the lie before it becomes a diagnosis.
- Chemical shift misplaces fat relative to water along one direction, leaving black/white outlines at fat–water borders (think kidney edges).
- Aliasing (wraparound) folds anatomy that pokes outside the field of view onto the opposite side of the image.
- Motion smears signal into repeating "ghosts" strung out in one specific direction.
- Susceptibility is metal (and air) wrecking the magnetic field — blooming black voids and warped, taffy-pulled anatomy, worst on gradient echo.
MRI is the most artistically gifted liar in the imaging department. CT mostly shows you what's there; MRI reconstructs a picture from radio signals and a lot of assumptions, and when those assumptions break, it doesn't shrug and give you static — it draws something crisp, plausible, and wrong. The whole skill of reading MRI is learning to spot when the scanner is making things up. Let's meet the four habitual fibbers.
Chemical shift: fat and water aren't on the same page
Fat and water protons sit in slightly different chemical neighborhoods, so they precess at slightly different frequencies. The scanner figures out where things are partly by frequency — so it nudges fat and water apart along the frequency-encoding direction. The result: at any sharp fat–water border (kidney against retroperitoneal fat, a vertebral body against its marrow), you get a bright band on one side and a dark band on the other, like a cheap color print where the ink layers didn't line up.
It looks like a real edge or even a thin lesion. It isn't. The giveaway: it sits exactly at a fat–water interface, it points along the frequency direction, and it flips sides if you swap the readout.
There are two flavors. The one above is chemical shift of the first kind (frequency misregistration). The second kind is the India-ink black outline around organs on opposed-phase gradient-echo images — that one is a feature we exploit on purpose, e.g. for adrenal adenomas and fatty liver. Same physics, opposite usefulness.
Aliasing: anatomy that won't stay in the box
You tell the scanner how big a field of view to image. If the patient is wider than that box — arms, hips, a generous belly — the part sticking out doesn't politely vanish. It gets wrapped around and pasted onto the opposite edge of the image. The medical name is aliasing; everyone calls it wraparound. Picture a panoramic photo stitched badly so the left edge of the scene reappears on the right.
So a hand resting beside the abdomen can show up ghosting through the liver. It's annoying because wrapped tissue can land squarely on the organ you care about and masquerade as pathology.
The fix is usually as dull as it sounds: enlarge the field of view, or turn on oversampling (no-phase-wrap). Knowing it's wraparound saves you from working up a "mass" that is just the patient's own arm visiting from across the image.
Motion: the ghosts in one direction
Anything that moves during the scan — breathing, a pounding aorta, swallowing, a kid who will not hold still — gets recorded inconsistently. MRI builds the image over time, so motion doesn't just blur; it produces discrete repeating copies, ghosts, marching across the image. Like trying to photograph a toddler at a slow shutter speed and getting four faint toddlers.
Here's the detail worth tattooing on your brain: ghosts propagate along the phase-encoding direction, regardless of which way the thing actually moved. That's because phase encoding is the slow axis, sampled across many repetitions, so it's the one that "remembers" change over time.
Pulsatile flow in a vessel throws ghosts straight across the image along the phase direction — and one of those ghosts can land inside an organ and pose as a lesion or a thrombus. Before you believe a "filling defect," ask whether it lines up with a pulsating vessel along the phase-encode axis. If it does, it's a ghost, not a clot.
Susceptibility: metal throws a tantrum
Different materials distort the magnetic field by different amounts — that's magnetic susceptibility. Tissue is fairly polite about it; metal absolutely is not. A surgical clip, a dental filling, or a hip prosthesis warps the local field so badly that protons nearby lose their signal entirely and the geometry around them gets stretched like taffy. You get a black blooming void with bright distorted edges, and anatomy that looks melted.
This is the same physics that makes susceptibility-weighted imaging so good at finding tiny bleeds and calcium — there we want the sensitivity. As an artifact, though, it can erase the exact spot you needed to see.
Crucially, susceptibility is far worse on gradient-echo sequences than on spin echo, because spin echo's refocusing pulse undoes much of the field-related signal loss. That difference is a tool: if a dark spot blooms on gradient echo but shrinks on spin echo, you're looking at susceptibility (blood, metal, calcium), not just low signal.
How to tell them apart at a glance
When something looks off, run the lineup. Each artifact has a tell — usually a direction and a location.
| Artifact | What it looks like | Dead giveaway |
|---|---|---|
| Chemical shift | Bright/dark bands at fat–water borders | Sits at a fat–water interface; along the frequency direction |
| Aliasing (wraparound) | Anatomy folded onto the opposite edge | Patient was wider than the field of view |
| Motion | Repeating ghosts smeared across the image | Propagates along the phase-encode direction |
| Susceptibility | Blooming black void, warped geometry | Worse on gradient echo; near metal/air |
The single most useful habit: when a finding looks too clean, too linear, or too perfectly aligned with an edge or a vessel, suspect the scanner before you suspect the patient. Real disease is usually messier than the artifact pretending to be it. (And if you're still shaky on why fat and water behave differently in the first place, it's worth a quick detour through MRI basics and tissue weighting.)