What you’ll be able to do after this lesson
- Name each of five core artefacts from its appearance and state its mechanism in one sentence
- Predict which encoding direction each one propagates along, and why
- Choose a first-line remedy that follows from the physics rather than from a memorised list
- Recognise when an artefact is diagnostically useful rather than a defect
Why artefacts are a physics topic
An artefact is not a fault. It is the encoding assumptions being violated, visibly. The scanner assumes that field variation is exactly what the gradients imposed, that everything precesses at the water frequency, that the object lies inside the field of view, and that nothing moves between phase-encode steps. Break any of those and the reconstruction produces something specific and predictable.
Which means the diagnostic skill is not memorising a table. It is asking: which assumption did this violate, and along which axis?
Aliasing (wrap)
Appearance: anatomy from outside the field of view folded onto the opposite side of the image, along the phase direction. Arms folded into an abdomen; a nose folded onto an occiput.
Mechanism: phase repeats every 360°. Tissue outside the FOV accumulates phase beyond one full cycle and is indistinguishable from tissue inside the FOV that accumulated the remainder. The reconstruction places it at the position it appears to belong to — the far side.
Why phase, not frequency: modern systems oversample and digitally filter the readout as a matter of course, so frequency-direction wrap is rarely seen. The phase axis has no such free remedy, because extra phase steps cost repetitions.
Remedies: enlarge the FOV along phase; enable phase oversampling (Siemens phase oversampling; GE “No Phase Wrap”; Philips foldover suppression), which acquires a larger phase FOV and discards the surplus, buying immunity with time; swap the phase direction to the narrower body axis; or place a saturation band over the offending anatomy.
Chemical shift
Appearance — type 1: a bright band on one side of a fat–water interface and a dark band on the other, along the frequency direction. Classically the rim of a kidney against perinephric fat, or the endplate margins on a spine.
Mechanism: fat and water protons resonate about 3.5 ppm apart — roughly 220 Hz at 1.5 T and 440 Hz at 3 T. Frequency encoding assumes every frequency difference is positional, so it maps fat to the wrong position along the readout axis. The misregistration widens with field strength and narrows as receiver bandwidth rises (since a wider bandwidth spans more hertz per pixel).
Appearance — type 2 (in-phase / opposed-phase): a black “India-ink” outline around every fat–water boundary, on all sides, on a gradient echo image at a specific TE.
Mechanism: because fat and water precess at different rates, their signals drift in and out of phase with each other. At 1.5 T the cycle is 1/220 Hz ≈ 4.5 ms, so they are opposed at about 2.2 ms and in phase at about 4.5 ms; at 3 T the cycle halves, giving roughly 1.15 ms and 2.3 ms. In a voxel containing both, the opposed-phase signals partially cancel. This is not a defect to eliminate — it is the basis of adrenal adenoma and hepatic steatosis assessment.
Remedies for type 1: raise the receiver bandwidth; apply fat suppression; swap the frequency direction so the shift falls somewhere harmless.
Susceptibility
Appearance: signal void larger than the object causing it, with geometric warping of surrounding anatomy and bright rims at the edges. Worst on gradient echo and EPI; worst near metal, air–tissue boundaries, and haemorrhage.
Mechanism: materials differ in magnetic susceptibility, so they distort the local field. Two things follow. Within a voxel, spins in different fields dephase — signal void. And across the image, the field is no longer what the gradients imposed, so positions are computed wrongly — geometric warping.
Remedies: use spin echo rather than gradient echo (the 180° pulse reverses static offsets — lesson 05); raise the receiver bandwidth (shorter readout, less time for error to accumulate); thinner slices; STIR or Dixon rather than spectral fat suppression, which fails when the field is distorted; and dedicated metal-artefact-reduction sequences where the scanner provides them.
When it is the finding: SWI and T2*-weighted gradient echo exist to exploit exactly this effect, for microhaemorrhage, iron and calcium. Same physics, opposite intent.
Anything concerning whether a specific implant may enter the scanner is a safety question governed by manufacturer labelling and local policy — see the site’s implant and safety guides, not this course.
Motion
Appearance: repeated ghost copies of moving structures, propagating along the phase direction regardless of which way the patient actually moved.
Mechanism: phase encoding is assembled from repetitions spread across the whole scan. Movement between them makes the lines mutually inconsistent, and the Fourier transform disperses that inconsistency into discrete ghosts. Periodic motion (respiration, pulsation) gives regularly spaced ghosts; random motion gives general blurring.
Remedies, in order: reduce the motion (coaching, comfort, immobilisation, a shorter sequence — an unrepeated good scan beats two rushed ones); gate or trigger to the physiology; suppress the source with a saturation band over moving fat or an inflowing vessel; apply flow compensation (gradient moment nulling) for pulsatile flow; use a motion-robust radial trajectory (Siemens BLADE, GE PROPELLER, Philips MultiVane, Canon JET), which re-samples the k-space centre with every blade; or relocate the ghosts by swapping the phase direction.
Note that the last option changes where the artefact lands, not whether it exists — and that averaging helps here for a reason unrelated to noise, by diluting inconsistent signal.
Truncation (Gibbs) ringing
Appearance: fine parallel bright and dark lines running alongside a sharp high-contrast boundary, decaying with distance. Most notorious as a false syrinx — a thin central line in the spinal cord — and at the skull–brain interface.
Mechanism: a sharp edge requires very high spatial frequencies to represent exactly. k-space is finite, so those frequencies are simply not measured, and the Fourier reconstruction of a truncated series overshoots and undershoots around the discontinuity. It is a mathematical property of finite sampling, not a scanner fault.
Remedies: increase the matrix along the affected direction (usually phase, where the matrix is lower), which pushes the ringing closer to the edge and reduces its amplitude; apply k-space filtering, at the cost of blurring; or simply recognise it. Ringing spaced regularly along the low-resolution axis, at a sharp boundary, in a cord that is otherwise normal, is the artefact — and identifying it correctly saves an unnecessary repeat and an unnecessary report.
At the console
An axial T2 of the cervical spine shows a thin bright line running the length of the cord, and the registrar asks whether it is a syrinx.
Work the physics. Which direction is it along? If it runs parallel to the cord and lies along the low-resolution phase axis, next to the sharp high-contrast CSF–cord boundary, truncation is the leading explanation. Test it the way the physics suggests: increase the phase matrix and repeat one slice. Real syrinx persists and follows anatomy across planes; ringing changes with the matrix and does not reproduce on a sequence with different resolution or a different phase direction. Confirm on the sagittal and on a sequence with a different encoding direction before anyone commits to a finding.
Second case, a rapid triage. A shoulder study shows a large dark region over the humeral head with warped anatomy at its margin, on an image where the fat is bright in some places and dark in others. Two artefacts, one cause. The dark warped region is susceptibility from an anchor; the patchy fat suppression is spectral fat-sat failing in the field distortion that same anchor is creating, compounded by the shoulder being off-isocentre. The physics-derived fixes point one way: move to spin echo, raise the bandwidth, replace spectral fat-sat with STIR, and get the anatomy closer to isocentre.
Check yourself
For each of the five artefacts, write down in one line: the mechanism, the direction it propagates along, and your first-line fix. Then find a recent case where you repeated a sequence, and ask whether the physics would have told you the repeat was futile.
This is a private prompt. Nothing you write is stored or assessed.
Key takeaways
- Artefacts are violated encoding assumptions. Ask which assumption broke, and along which axis.
- Aliasing folds out-of-FOV tissue along phase because phase repeats every 360°. Oversampling, a larger phase FOV or a swapped direction are the fixes.
- Chemical shift misregisters fat by ~3.5 ppm (≈220 Hz at 1.5 T, ≈440 Hz at 3 T) along frequency; higher bandwidth narrows it. Opposed-phase cancellation is a separate, diagnostically useful effect.
- Susceptibility both dephases (void) and misplaces (warping). Spin echo, high bandwidth, thin slices and STIR/Dixon mitigate it — and SWI exploits it deliberately.
- Motion ghosts propagate along phase whatever the direction of movement, because phase is assembled over time. Reduce, gate, suppress, or relocate.
- Truncation ringing is finite sampling at a sharp edge, not a fault. Increase the matrix along the affected axis, and recognise the false syrinx before repeating.
References
- Zaitsev M, Maclaren J, Herbst M. Motion artifacts in MRI: a complex problem with many partial solutions. Journal of Magnetic Resonance Imaging. 2015;42(4):887–901 https://doi.org/10.1002/jmri.24850
- Hargreaves BA, Worters PW, Pauly KB, Pauly JM, Koch KM, Gold GE. Metal-induced artifacts in MRI. American Journal of Roentgenology. 2011;197(3):547–555 https://doi.org/10.2214/AJR.11.7364
- Elster AD. Questions and Answers in MRI — ‘Fat-water chemical shift’, ‘Gibbs ringing’, ‘Aliasing’ https://mriquestions.com/f-w-chemical-shift.html
- Radiopaedia. MRI artifacts (physics articles: aliasing, chemical shift, susceptibility, truncation) https://radiopaedia.org/articles/mri-artifacts
- HighYield Imaging. High-Yield MRI Trainee Manual — artefacts and fixes reference table https://highyieldimaging.com/mri-trainee-manual/