MRI Artefacts — Master Page: Classification, Physical Origins, Recognition and Management
MRIninja Knowledge Base | Master / Reference Page — MRI Artefacts Version 2.1 — October 2026
Builds on: MRI Physics — Fundamentals and Principles · Signal Localisation and Image Formation (k-Space) · Relaxation Phenomena · RF Pulses and Pulse Sequence Physics · Flow, Motion, and Diffusion Physics · Chemical Shift and Fat/Water Physics · MRI Parameters — Overview and Classification
1. Scope, Status of This Page and How to Use It
This page is the hub for MRI artefacts on MRIninja. It replaces the previous version 1.0 (Physics of MRI Artefacts, a child page of the physics master) and keeps its verified physics for five artefacts (Section 4). Its job is different from a child page: it does not try to describe every artefact in depth. It defines what is meant by an artefact, gives one classification that every later page can use, states the physical principles that are shared by most artefacts, gives a single table of all the artefacts the site intends to cover (Section 6), and gives the decision aids that are common to all of them (Sections 7 and 8). Each artefact then gets its own child page (Section 12), which will assume everything written here.
Artefact content that already exists elsewhere on the site is not repeated and is linked in the taxonomy: pulsatile ghosting, intravoxel dephasing and flow misregistration (Flow, Motion, and Diffusion Physics, Section 4), inflow and outflow effects (Inflow and Outflow Effects), chemical shift physics (Chemical Shift and Fat/Water Physics, Chemical Shift — Theoretical Foundations), phase oversampling and anti-aliasing (Phase Oversampling, Antialiasing Options and Strategies, Fold-Over Direction), maximum intensity projection artefacts (MIP in Angiography) and motion correction by radial blades (MultiVane and the Philips “Motion-Free” Family).
Conventions. “Artefact” is spelled in the British form on this page. Statements that are the author’s derivation or inference rather than a cited finding are labelled Reasoning. Figures 1–3 are computed from the stated models and are not measurements. A numbered citation [n] means that the cited source was read at least at abstract level; where only an abstract was available the source is flagged in the reference list.
2. Definition and Classification Framework
2.1 Definition Used on MRIninja
An MRI artefact is any feature of the image that does not correspond to the true spatial distribution of the tissue property being imaged. Three groups are worth separating because they call for different responses.
- Reconstruction-consistent artefacts — the signal is real but is assigned to the wrong place or is missing: displacement, replication, wrap-around, distortion, signal void. The mislocalisation follows from how position is encoded (Section 3).
- Acquisition-inconsistency artefacts — the object changes while k-space is being filled (motion, pulsatile flow, signal drift, RF interference), so the Fourier reconstruction receives data that do not belong to one object.
- Physical-signal artefacts — the signal itself differs from what the reader expects because of a real physical effect: magic angle, inflow enhancement, susceptibility-related dephasing, dielectric shading, supplemental-oxygen effects on FLAIR. Nothing is mislocalised; the interpretation is what goes wrong.
2.2 Four Axes for Describing Any Artefact
Every artefact page on the site will state the same four attributes so that pages can be compared and searched.
| Axis | Question | Typical answers |
|---|---|---|
| Origin | Where does the cause sit? | Patient (motion, implant, tissue property) · sequence and parameters · reconstruction · scanner hardware (magnet, gradients, RF, coil) · environment |
| Image effect | What does it do to the image? | Replication (ghost, wrap) · displacement · signal void or loss · signal pile-up · ringing · intensity shading · geometric distortion · noise pattern (line, spike, streak) · false signal |
| Direction dependence | Along which axis does it act? | Phase-encoding · frequency-encoding · slice-selection · none (global). This decides whether swapping the phase- and frequency-encoding directions moves the artefact (Section 7). |
| Controllability | Can the operator change it? | Operator parameter · sequence choice · patient preparation · service or physicist intervention · not removable (physics): recognise, do not “fix” |
3. Shared Physical Principles
Reasoning: most of the artefacts in Section 6 follow from five statements. They are the reason a taxonomy of this size is manageable: each child page reduces to one or two of these principles plus the specific cause.
3.1 Position Is Encoded in Frequency and Phase
The scanner assumes that precession frequency depends on position only through the applied gradients, and that phase depends on position only through the phase-encoding gradient (k-space page). Anything else that changes the precession frequency of a spin — susceptibility differences, chemical shift, main-field inhomogeneity, gradient error — is read as a change of position. Along the frequency-encoding axis the displacement in pixels equals the frequency offset divided by the receiver bandwidth per pixel (Figure 3). In a conventional Cartesian sequence each phase-encoding line is acquired after its own excitation, so an off-resonance spin adds the same extra phase to every line and the displacement appears along the frequency-encoding axis. In echo-planar imaging successive lines follow one another within a single readout train, so the off-resonance phase accumulates from line to line and the displacement appears along the phase-encoding axis, where the effective bandwidth per pixel is low and the shifts can therefore be large.
3.2 Reconstruction Assumes One Unchanging Object
The Fourier transform treats every k-space sample as a measurement of the same object. When the object (or the signal it contributes) changes between phase-encoding lines — respiration, cardiac pulsation, bulk movement, flow with variable velocity, a drifting scanner — the error is spread across the image, and in a Cartesian acquisition it is spread along the phase-encoding axis, because that axis is the one sampled over the longest time (Flow, Motion, and Diffusion Physics, Section 4.1). Readouts that sample k-space in other orders (radial, spiral, blades) redistribute the error differently (MultiVane page).
3.3 Sampling Is Discrete and Finite
Discrete sampling gives aliasing when the object extends beyond the field of view (Section 4.2); finite extent of the sampled k-space gives truncation ringing at sharp edges (Section 4.3). Both are properties of the Fourier relationship, not of a particular scanner.
3.4 RF, Gradient and Receive Fields Are Not Perfectly Uniform
The transmitted RF field (B1), the gradient fields and the receive-coil sensitivity all deviate from the ideal. Consequences are intensity shading, flip-angle errors, geometric distortion growing with distance from the isocentre, and local signal drop-outs. These are properties of the hardware and of the patient’s own electrical properties at the Larmor frequency, which is why they become more prominent at higher field strength (High-Field Physics) and why they are a hardware-quality topic (Magnetic Field Hardware and Homogeneity) as much as a protocol topic. Zhuo and Gullapalli organise MR artefacts by the scanner component they arise from — main magnet, gradient coils, RF transmitter and receiver, and the reconstruction algorithm [4].
3.5 Some “Artefacts” Are Real Signal Behaviour
The magic angle effect, inflow enhancement, susceptibility-related T2* shortening and relaxation changes from supplemental oxygen are real physical signal changes. They cannot be removed by changing a parameter in the way aliasing can; they are managed by knowing when they occur and by changing the conditions (echo time, orientation, sequence family) to show that the finding depends on them.
4. Physical Origin of Five Core Artefacts (Retained from Version 1.0)
The text of this section is the verified content of version 1.0 of this page. One quantitative nuance was added in Section 4.3 and is marked there. Flow- and motion-related artefacts are in Flow, Motion, and Diffusion Physics and are not repeated.
4.1 Susceptibility Artefact
4.1.1 Physical Origin — Magnetic Susceptibility Variation
Every material — including every tissue type in the human body, and every object placed near the scanner — has a magnetic susceptibility: a physical property describing how strongly it becomes magnetised in response to an external field. Schenck’s comprehensive 1996 review formalised the practical classification of MRI-relevant susceptibility behaviour, distinguishing materials whose susceptibility is close enough to that of tissue and water to produce negligible field distortion from those whose susceptibility differs enough to produce clinically significant local field distortion [1]. Because different tissue types (and, most dramatically, interfaces between tissue and air, tissue and bone, tissue and metal, or tissue and haemorrhage/calcification products) have measurably different magnetic susceptibility, the local magnetic field near such interfaces is distorted from its nominal, intended value — directly perturbing the local Larmor frequency, as introduced in Relaxation Phenomena (T2′ and field inhomogeneity), but now considered as a spatially localised, interface-specific phenomenon rather than a general contributor to T2* decay.
4.1.2 Consequences — Signal Loss and Geometric Distortion
Susceptibility-related field distortion produces two distinct, simultaneously occurring practical effects. First, the accelerated local dephasing described in Section 4.1.1 produces localised signal loss, most pronounced on T2*-weighted gradient-echo sequences (which, as established in RF Pulses and Pulse Sequence Physics, have no refocusing pulse to actively reverse this field-inhomogeneity-driven dephasing) and correspondingly much less pronounced on spin-echo-family sequences. Second, because spatial position is assigned based on local Larmor frequency, the same local frequency shift that causes signal loss also causes genuine geometric mislocalisation — pixels near a strong susceptibility interface are displaced from their true anatomical position — a distinct but physically related consequence of the same underlying field distortion. Hargreaves and colleagues describe the same set of consequences for metal implants: signal loss, failed fat suppression, geometric distortion and bright signal pile-up, all arising from large changes in resonant frequency [7].
4.2 Aliasing / Wrap-Around Physics
4.2.1 The Sampling Theorem and the Nyquist Criterion
Any system that samples a continuous signal at discrete intervals — exactly the situation in MRI, where k-space is sampled at discrete points rather than continuously (k-space page) — is subject to a fundamental mathematical constraint first rigorously established in the general signal-processing literature by Harry Nyquist in 1928: a signal can be unambiguously reconstructed from its samples only if it is sampled at a rate at least twice its highest frequency component [2]. Applied to MRI’s specific spatial-frequency (k-space) sampling, this becomes a direct geometric statement: the k-space sampling interval must be no coarser than 1/FOV (field of view) along a given encoding direction, or spatial frequencies originating from anatomy actually located outside the nominal field of view become mathematically indistinguishable, after Fourier transformation, from spatial frequencies properly originating from within it.
4.2.2 Wrap-Around as a Direct Consequence
When the Nyquist criterion described in Section 4.2.1 is violated — anatomy extends beyond the encoded field of view along the phase-encode or frequency-encode direction — the reconstructed image does not simply crop or blank the excess anatomy; instead, the excess anatomy’s signal is mathematically indistinguishable from signal originating from within the field of view, and appears wrapped around to the opposite edge of the image, superimposed on genuinely intended anatomy. Because the phase-encode direction is sampled discretely, one line at a time, aliasing is generally more prominent along phase-encode than frequency-encode, where oversampling the continuously-sampled readout signal beyond the nominal Nyquist rate is comparatively cheap and is, in practice, done routinely — the direct physical reason frequency-encode-direction aliasing is rarely a clinical problem while phase-encode-direction aliasing remains a genuine, actively-managed concern, addressed practically (phase oversampling, appropriately sized field of view) in Phase Oversampling and Antialiasing Options and Strategies. Krupa and Bekiesińska-Figatowska note in their review that aliasing can be removed by enlarging the field of view [6].
4.3 Truncation (Gibbs) Ringing
4.3.1 Physical and Mathematical Origin
Because any real, finite acquisition samples only a finite region of k-space rather than the theoretically infinite extent that would be required to represent a perfectly sharp image edge exactly (per the Fourier-transform relationship between k-space and image space), the reconstructed image of any genuinely sharp intensity transition — most classically, at high-contrast tissue boundaries such as the spinal cord/CSF interface or the boundary of a syrinx — necessarily shows a characteristic pattern of parallel, alternating bright and dark bands adjacent to the true edge, a direct mathematical consequence of abruptly truncating an infinite Fourier series at a finite number of terms, a phenomenon mathematically identical (though applied here to two-dimensional spatial data rather than the one-dimensional mathematical series it was first described for) to the general Gibbs phenomenon in Fourier analysis. In the spinal cord a truncation band can mimic a syrinx [6].
4.3.2 Determinants of Severity (revised)
Truncation ringing’s prominence depends on how sharp the transition is relative to the pixel and on how much contrast it carries: a larger abrupt intensity transition requires more high-spatial-frequency k-space content to represent sharply (k-space page). Nuance added in version 2.0. In the idealised band-limited model the fractional overshoot at an ideal edge is about 9% of the edge height and is the same for every matrix; what changes with matrix is the spacing of the rings, which is proportional to the pixel size (Figure 1). Version 1.0 stated that ringing is more pronounced with coarser resolution; in the model this is true of its spatial extent and of its conspicuity (rings several millimetres wide can resemble a lesion or a syrinx), not of its fractional amplitude. Real images are further modified by zero-filling, apodisation filters and partial Fourier acquisition, so the model gives the scaling, not the amplitude seen on a given scan. This is the direct physical reason truncation artefact is a well-recognised, specifically-named pitfall in spinal cord imaging (where high cord/CSF contrast combines with comparatively modest in-plane resolution).
Figure 1 — Truncation ringing at an ideal edge for three pixel sizes
4.4 RF-Related Artefacts — the Zipper Artefact
Because the receiver is tuned to detect signal specifically at and immediately around the expected Larmor frequency (see Fundamentals of Nuclear Magnetic Resonance), any stray, unintended RF energy at or near that frequency — from imperfect RF shielding of the scan room, electronic interference from nearby equipment, or an unintentional resonance within the scanner’s own RF chain — is detected by the receiver exactly as though it were genuine MR signal. Because such stray RF energy is typically present at a single, narrow frequency rather than distributed smoothly across the expected signal bandwidth, it appears in the reconstructed image as a characteristic thin line of alternating bright and dark pixels running along the phase-encode direction, at the specific spatial position corresponding to its interference frequency — the “zipper” artefact, a direct, diagnostic-quality indicator of an RF shielding or interference problem rather than a normal, expected imaging phenomenon (distinguishing it from the other artefacts in this section, all of which occur routinely even with fully functioning hardware).
4.5 The Magic Angle Effect
The magic angle effect has a genuinely distinct physical origin from every other artefact in this section: it arises from the anisotropic nature of dipolar coupling in highly organised, collagen-rich tissue (tendons, ligaments, and similar structures), where water molecules are tightly bound in a restricted geometric relationship to collagen fibrils. Erickson and colleagues’ original 1991 clinical description demonstrated that when such a structure is oriented at approximately 54.7° (the “magic angle,” the specific angle at which the standard dipolar coupling term (3cos²θ − 1) equals zero) relative to B0, this dipolar coupling’s normally dominant contribution to rapid T2 decay is eliminated, causing an artefactual T2 lengthening and correspondingly increased signal on short-TE sequences — signal that can closely mimic the appearance of genuine tendinopathy or partial tear [3]. Because this effect depends on a genuine, real physical interaction (dipolar coupling) rather than a hardware or sampling artefact, it is not eliminated by any hardware or acquisition parameter adjustment in the way aliasing or truncation ringing can be — it is instead recognised and distinguished from true pathology primarily by its characteristic dependence on both anatomical orientation and echo time (the artefact diminishes or disappears on longer-TE sequences), discussed practically in the relevant musculoskeletal anatomical protocol pages elsewhere on MRIninja.
5. Two Quantitative Illustrations That Apply Across Families
Figure 1 (Section 4.3) quantifies truncation. Two further quantities recur in almost every later child page and are therefore given once here: where a motion ghost falls, and how far an off-resonance signal is displaced.
5.1 Where a Periodic-Motion Ghost Falls
Reasoning (derivation). In a Cartesian acquisition with one phase-encoding line per repetition and one signal average, a motion of frequency f modulates successive lines by f·TR cycles per line. The Fourier transform along the phase-encoding axis turns that modulation into a displaced replica: with Npe phase-encoding steps the replica is displaced by Npe·f·TR pixels, equivalent to the fraction frac(f·TR) of the field of view, wrapped around it. Ghosts occur on both sides of the structure, so the offset can be folded to 0–0.5 FOV (Figure 2). Two consequences are practical. First, the ghost position is set by the product of motion rate and TR, not by the anatomy, so changing TR moves the ghost; this is a way to move a ghost off a region of interest and a way to confirm that a replica is motion-related. Second, when f·TR is a whole number the ghost falls on the structure itself. The model does not apply as written to turbo spin echo, multi-shot or radial readouts, to averages acquired in sequence, or to non-periodic motion, for which ghosts smear along the phase-encoding axis.
Figure 2 — Ghost offset versus rate of periodic motion
5.2 How Far an Off-Resonance Signal Is Displaced
Reasoning (derivation). A frequency offset Δf is displaced along the frequency-encoding axis by Δf divided by the receiver bandwidth per pixel (BWpx = total bandwidth / number of frequency-encoding samples). With Δf = ppm × 42.577 MHz/T × B0 (Figure 3); the fat–water case, with worked values at 1.5 T and 3 T, is held in Chemical Shift Artefact of the First Kind (Section 4). Because displacement in millimetres equals pixels × pixel size and BWpx = BW/Nx, the displacement in millimetres equals Δf × FOV / BW. At fixed field of view and fixed total receiver bandwidth, the displacement in millimetres does not depend on the matrix; it falls only when total bandwidth rises (with an SNR cost; Acquisition Bandwidth) or when the offset is removed (fat suppression, spin-echo refocusing, shorter echo time). The same relation governs susceptibility- and metal-related displacement, where Δf can be much larger than the fat–water value [7]. The 1 and 5 ppm curves in Figure 3 are arbitrary illustrative magnitudes.
Figure 3 — Displacement in pixels versus receiver bandwidth per pixel
6. Master Taxonomy of MRI Artefacts
The tables list the artefacts for which a dedicated page is planned, grouped in nine families with a short introduction each. Each row gives the typical appearance, the mechanism in one line, the main levers and what is already on the site. “Existing coverage” means that the mechanism is explained somewhere on the site, not that an artefact-specific page exists. Proposed IDs in the range 9601–9699 are provisional reservations (Section 12). The list was compiled from the physics in Sections 3–5, the technical reviews [4–8] and common clinical practice; it is probably still incomplete, and rows are added as child pages are researched. Where the mechanism or remedy is a general-knowledge summary that has not been checked against a source on this page it is marked “(to be sourced)”.
A. Patient motion and physiology
Common cause: the object changes while k-space is being filled (Section 3.2). Periodic motion gives discrete ghosts whose position depends on the motion rate and TR (Figure 2); non-periodic motion gives smearing and blurring. The family differs mainly in the motion source and therefore in the remedy.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Bulk voluntary and involuntary motion | Ghost replicas and blurring along the phase-encoding axis; whole image degraded | Inconsistent phase-encoding lines (3.2) | Coaching, immobilisation, sedation where appropriate, shorter or motion-robust acquisition, swap phase direction | 9207 · 9028 | 9601 |
| Respiratory motion | Ghosts of anterior abdominal or chest wall fat and of organs; blurring of liver and lung bases | Periodic motion at 10–25 cycles/min (Figure 2) | Breath-hold, triggering or navigators, saturation bands, radial readout | 9207 · 9586 | 9602 |
| Cardiac and arterial pulsation | Ghost chain from vessels (aorta, carotids, venous sinuses) along phase encoding | Signal change locked to the cardiac cycle | Flow compensation, triggering, saturation bands, phase direction | 9207 | 9603 |
| CSF pulsation | Signal loss or ghosting in CSF spaces; mimics lesion in spinal canal and posterior fossa | Pulsatile CSF flow dephases and refreshes signal | Flow compensation, gradient moment nulling, sequence choice, orientation | — | 9604 |
| Swallowing, tongue and neck motion | Blurring and ghosting in neck, thyroid and cervical spine | Short irregular motion (non-periodic) | Instruction, shorter acquisition, saturation band over airway, phase direction | — | 9638 |
| Eye movement and blinking | Ghosts of the globes across the brain or orbit along phase encoding | Moving high-signal structures; phase-encoding direction determines where ghosts fall | Eyes closed or fixation, phase direction, saturation band | — | 9639 |
| Peristalsis and bowel gas | Blurring and ghosts in abdomen and pelvis; focal signal loss from gas | Irregular non-periodic motion; gas–tissue susceptibility | Fasting where applicable, antiperistaltic agent per local protocol, shorter acquisition, saturation band | 9586 | 9637 |
| Fetal and neonatal motion | Unpredictable blur; slice-to-slice misalignment | Spontaneous movement during acquisition | Single-shot acquisitions and slice-wise registration (to be sourced) | — | 9640 |
| Pulsation and motion signal loss in diffusion-weighted imaging | Focal signal drop-out (e.g. left lobe of the liver, brain parenchyma) | Motion during the strong diffusion gradients adds unpredictable phase | Triggering, shorter diffusion encoding, repeat or average (to be sourced) | 9012 · 9013 | 9641 |
| Transient severe motion after gadoxetate | Marked respiratory-motion degradation of the arterial phase in a subset of patients | Mechanism debated; reported with gadoxetate (to be sourced) | Motion-robust or multi-arterial-phase acquisition (to be sourced) | 9101 | 9642 |
B. Flow and angiography
Moving blood changes signal through inflow of unsaturated spins, outflow of excited spins, velocity-dependent phase and intravoxel phase spread (9023, 9207, 9213). The artefacts therefore depend on the sequence family more than on the vessel.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Inflow enhancement and entry-slice effect | Bright vessel signal at the first slices of a stack or slab; mimics thrombus or enhancement | Fresh spins entering the slice | Saturation, slice orientation, flip angle, gradient-echo vs spin-echo | 9023 · 9213 | 9616 |
| Outflow and time-of-flight signal loss; slow flow | Absent or reduced flow signal in a patent vessel | Excited spins leave the slice; slow flow saturates | Shorter TE, thinner slabs, ramped flip angle | 9023 · 9024 · 9025 | 9616 |
| Flow voids, intravoxel dephasing and flow misregistration | Signal void at stenosis or aneurysm; displaced vessel along phase encoding | Wide velocity spread in a voxel; spin moves between encoding and readout | Flow compensation, shorter TE, higher resolution | 9207 | 9617 |
| Phase-contrast velocity aliasing and background phase errors | Abrupt velocity sign reversal; offset in flow quantification | VENC below peak velocity; eddy-current and concomitant-gradient phase offsets | Higher VENC, background-phase correction, unwrapping | 9022 | 9643 |
| Slab-boundary (venetian-blind) artefact in 3D multi-slab TOF | Periodic intensity bands across stacked slabs | Flip-angle profile and saturation across overlapping slabs | Slab overlap, ramped excitation, slab thickness | 9024 · 9025 · 9599 | 9644 |
| Contrast-enhanced MRA timing artefacts | Ringing at vessel edges, ghosting, venous contamination | Bolus arrives away from the k-space centre acquisition | Test bolus, bolus tracking, centric ordering | 9101 | 9645 |
| T1 shine-through and venous contamination in MRA | Bright non-vascular tissue on MIP | Short-T1 tissue (fat, haematoma) bright on gradient echo | Fat suppression, subtraction, source-image review | 9024 · 9027 | 9616 |
| Arterial spin labelling artefacts | Low perfusion signal; labelling-plane or transit-time errors | Labelling efficiency and arterial transit time; susceptibility | Post-labelling delay, labelling-plane placement | 9021 · 9301 | 9646 |
C. Sampling, reconstruction and display
These artefacts are properties of the data processing chain (Section 3.3) rather than of the patient. They include the finite and discrete sampling effects and the newer reconstruction methods whose artefact patterns are less well described.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Aliasing / wrap-around (phase, frequency and slice directions) | Anatomy beyond the field of view folded onto the opposite edge; slab wrap in 3D | Nyquist violation (4.2) | Phase oversampling, larger FOV, saturation bands, slice oversampling | 9585 · 9594 · 9593 · 9595 | 9611 |
| Truncation (Gibbs) ringing; syrinx pitfall | Parallel bright and dark lines beside sharp edges; spinal cord pseudo-syrinx | Finite k-space (4.3) | Larger matrix, filtering, avoid aggressive partial Fourier | Section 4.3 | 9610 |
| Parallel-imaging noise (g-factor) and residual aliasing | Noise increase in the image centre; residual fold-over | Ill-conditioned unfolding; calibration mismatch | Lower acceleration factor, better coil geometry, recalibration after motion | 9588 | 9615 |
| Compressed-sensing and deep-learning reconstruction artefacts | Over-smoothing, loss of small detail, plausible-looking structures that are not present | Prior-based reconstruction (to be sourced) | Reduce acceleration, compare with conventional reconstruction | 9212 | 9630 |
| Partial Fourier, zero-filling, interpolation and filter effects | Blur, ringing, apparent increase in resolution without new detail | Post-acquisition modification of k-space or image | Know the native acquisition matrix | 9587 | 9629 |
| Moiré (fold-over) fringes in gradient echo | Fringe patterns where wrapped signal overlaps the object | Wrapped and in-field signal of differing phase interfere (to be sourced) | Prevent wrap; larger FOV | — | 9631 |
| Central point (zero-frequency) artefact | Bright dot or short line at the image centre | Offset in the receiver data (to be sourced) | Service intervention if persistent | — | 9647 |
| Receiver overload and data clipping | Banding, ghosting or abnormal noise after a gain or prescan error | Analogue-to-digital converter saturation | Repeat prescan; service check | — | 9648 |
| Radial, spiral and blade undersampling streaks | Fine streaks radiating from bright structures instead of ghosts | Undersampled non-Cartesian k-space | More spokes or blades, streak-reduction reconstruction | 9028 | 9649 |
| Stitching artefacts in multistation imaging | Intensity steps and misalignment at station junctions | Gradient non-linearity and B0 variation at station edges | Overlap, distortion correction, normalisation | — | 9650 |
| Partial-volume averaging | Apparent lesion or loss of a small structure | Slice thickness and voxel size larger than the structure | Thinner slices, isotropic 3D, orthogonal planes | 9583 | 9651 |
| Reformat and projection artefacts (MIP, MPR, volume rendering) | Stair-steps, false stenosis or occlusion on reformats | Rendering of the data, not of the acquisition | Source-image review, thin slices | 9027 | 9652 |
| Subtraction misregistration | Edge artefact on subtracted dynamic images | Movement between mask and post-contrast series | Co-registration, motion control | 9101 | 9628 |
D. Chemical shift and fat
The 3.5 ppm fat–water difference is a frequency offset and therefore behaves like any off-resonance (Section 5.2); it also determines which echo times put fat and water in or out of phase. Fat suppression adds a second set of failure modes.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Chemical shift, first kind (misregistration) | Bright and dark rim at fat–water interfaces along the frequency-encoding axis | Frequency offset read as position (Figure 3) | Higher bandwidth, fat suppression, swap encoding axes | 9206 · 9596 · 9597 | 9607 (published) |
| Chemical shift, second kind (India-ink / opposed phase) | Dark outline at fat–water boundaries on opposed-phase gradient echo | Fat and water in one voxel out of phase at the TE | Choose in-phase TE; understand diagnostic use | 9206 | 9608 (published) |
| Chemical shift in the slice-selection direction | Fat and water slices displaced relative to each other | Frequency offset during slice selection | Higher slice-selection bandwidth, thicker slice | — | 9653 |
| Fat-suppression failure | Patchy or absent fat suppression, uneven brightness | Field and B1 non-uniformity; wrong method for the region | Different suppression method, shimming, centring | 9006 | 9609 |
| Incomplete nulling in STIR and FLAIR | Residual CSF or fat signal; suppression of other short-T1 tissues | Inversion error (B1), inflow, T1-dependence | Adiabatic inversion, correct TI, flow management | 9006 · 9007 · 9020 | 9654 |
| Dixon fat–water swap | Fat and water labelled inversely in part of the image | Ambiguity in phase estimation | Better shimming, robust algorithm | — | 9632 |
E. Susceptibility and metal
Local field distortion gives signal loss, geometric displacement and failed fat suppression at the same time (4.1, [7]). Gradient-echo sequences are the most sensitive; spin echo reduces but does not remove the problem.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Susceptibility (air–bone, blood products, calcification) | Signal loss with distortion at skull base, sinuses, bowel gas, haemorrhage | Local field distortion (4.1) | Spin echo, shorter TE, higher bandwidth, thinner slices | 9011 | 9606 |
| Metal implant artefacts | Large void with bright pile-up rim, distortion, failed fat suppression | Very large off-resonance [7] | Orientation, bandwidth, spin echo, metal-reduction sequences, lower field | 9004 | 9605 |
| Dental and orthodontic hardware | Facial and skull-base signal void, distorted sinuses and orbits | Ferromagnetic or conductive hardware near the head | Sequence choice, orientation, removal of removable items | — | 9655 |
| Blooming and phase wrap on gradient echo, SWI and QSM | Enlarged dark foci; phase discontinuities | Dephasing extends beyond the object; phase aliasing | TE, resolution, phase unwrapping | 9011 · 9016 | 9624 |
| Foreign bodies and patient-related sources | Local void or distortion from tattoos, make-up, clothing, hair products | Ferromagnetic or conductive material [6] | Screening, removal, documentation | — | 9633 |
| Concentrated gadolinium T2* effects | Signal loss in renal collecting system or bladder in excretory phases | High local contrast concentration shortens T2* (to be sourced) | Timing, sequence choice | 9101 | 9656 |
| Supplemental-oxygen effect on FLAIR | Subarachnoid hyperintensity on FLAIR | Dissolved oxygen shortens CSF T1 [6] | Avoid oxygen where possible; know the pattern | 9007 | 9626 |
F. Echo-planar and diffusion imaging
Echo-planar readouts acquire many lines after one excitation, so their effective bandwidth per pixel along phase encoding is very low (3.1). Combined with strong diffusion gradients, this family has its own set of ghosts, distortions and quantitative errors.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| EPI Nyquist (N/2) ghost | Half-field-of-view ghost along phase encoding | Mismatch of odd and even lines (alternating readout polarity) | Reference-scan correction, centring, service check | 9010 | 9612 |
| EPI geometric distortion and fat displacement | Stretching, compression, pile-up; displaced fat if unsuppressed | Off-resonance accumulates between lines | Parallel imaging, shorter echo spacing, reversed phase-encoding correction | 9010 · 9014 | 9613 |
| Eddy-current artefacts | Image shear, scaling and shift that depend on gradient direction | Induced fields after strong gradient pulses | Pre-emphasis, twice-refocused diffusion encoding, post-processing registration | 9208 | 9614 |
| T2 shine-through and T2 blackout | Bright DWI signal without restricted diffusion; dark DWI in short-T2 lesions | DWI signal depends on T2 and ADC | Interpret with ADC and T2 | 9012 · 9013 | 9627 |
| Diffusion b-value and ADC bias from gradient non-linearity | ADC values that vary with position in the magnet | Spatially varying effective b-value (to be sourced) | Correction, central positioning | 9208 | 9657 |
G. RF, receive chain and field uniformity
The transmit field, the receive sensitivity and the main field are never perfectly uniform (3.4). Some effects are properties of the patient and field strength; others are signs of hardware faults that need engineering investigation.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Zipper artefact and RF interference | Thin line of alternating bright and dark pixels at a fixed image position | Narrow-band extraneous RF (4.4) | Close door, remove electronic items, service call | Section 4.4 | 9621 |
| k-space data spikes (herringbone, corduroy) | Oblique line pattern across the whole image | Isolated corrupted k-space points (to be sourced) | Repeat; service if recurrent | — | 9621 |
| B1 inhomogeneity and dielectric/standing-wave shading | Central brightening and peripheral darkening at 3 T | Wavelength effects in the body; flip-angle error | Dielectric pads, parallel transmission, intensity correction | 9211 · 9208 | 9618 |
| Receive-coil intensity bias, element drop-out and loading errors | Smooth shading toward the coil; patchy drop-out | Sensitivity profile not corrected; faulty element | Intensity normalisation, check coil connection | — | 9619 |
| Slice cross-talk, cross-excitation and slice-profile effects | Reduced contrast or signal from adjacent slices | Imperfect slice profiles excite neighbouring slices | Slice gap, interleaving | 9591 · 9583 | 9623 |
| Poor shimming and off-centre imaging | Failed fat suppression, distortion, banding, signal loss at the periphery | B0 non-uniformity over a large or off-centre volume | Centring, shim volume, smaller FOV | 9208 | 9658 |
H. Gradient and scanner stability
Gradient fields deviate from linearity away from the isocentre, and mechanical or thermal instability can change the field during a study. These effects are small for detail and large for measurement.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Gradient non-linearity distortion | Geometric distortion growing with distance from isocentre | Gradient field deviates from linear [5] | Distortion correction, centring | 9208 | 9620 |
| Vibration, B0 drift and table motion | Banding or blur that varies in time | Instability of scanner or environment | Service check | 9208 | 9634 |
| Quality assurance for artefact detection | — | Phantom tests, acceptance and routine checks [4] | Defined escalation pathway | — | 9635 |
I. Tissue-property and sequence-specific
Some of the most frequent pitfalls depend on a particular sequence family rather than on hardware: the signal is genuine but the sequence changes what is seen.
| Artefact | Typical appearance | Mechanism | Main levers | Existing coverage | Planned page |
|---|---|---|---|---|---|
| Magic angle effect | Increased tendon signal at about 55° to B0 on short TE | Dipolar coupling in collagen (4.5) | Longer TE, reposition | Section 4.5 | 9622 |
| bSSFP banding | Dark bands across the image | Off-resonance in steady-state free precession | Shorter TR, shimming, phase cycling | 9015 | 9625 |
| FLAIR CSF-flow and vessel artefacts | CSF and vascular hyperintensity not due to pathology | Inflow of unnulled signal; pulsation | Sequence settings, 3D acquisition | 9007 | 9626 |
| Turbo spin-echo blurring and bright fat | Blur along phase encoding; bright fat | T2 decay along the echo train; reduced J-coupling effect (to be sourced) | Shorter echo train, echo spacing | 9005 | 9659 |
| Cardiac-MR artefact set | Cine and late-enhancement degradation from gating and arrhythmia; banding | Trigger errors; off-resonance | ECG quality, trigger delay, shimming | — | 9636 |
| MR spectroscopy artefacts | Wrong voxel contents; lipid contamination | Chemical-shift displacement of the voxel; out-of-voxel signal (to be sourced) | Voxel placement, outer-volume suppression | — | 9660 |
Limits of this table. The families overlap by design (for example, EPI distortion is both a sequence-specific and an off-resonance artefact). The table is a navigation aid built from the physics in Sections 3–5, the technical reviews [4–8] and the pages already on the site; mechanism lines for artefacts not discussed in Sections 4–5 are summary statements that each child page will source individually. It is not a published classification, and none exists as a consensus standard. Rows marked “—” under existing coverage were checked by text search of this site on 18 October 2026; absence of a hit does not exclude a short mention. Entries sharing a planned ID will be covered by the same child page.
7. Recognition: From Appearance to Mechanism
Reasoning: the discriminating tests below follow from the mechanisms in Sections 3–5. They are standard practice in the sense that they are the changes that a mechanism predicts will move, change or abolish the finding; true anatomy and pathology do not follow them. Only a change that the mechanism predicts counts as evidence.
| Appearance | First hypotheses | Discriminating change | Expected if artefact |
|---|---|---|---|
| Replica of a structure, displaced along the phase-encoding axis, often smeared | Periodic or non-periodic motion; pulsatile flow | Swap phase- and frequency-encoding directions; change TR; add saturation band; use triggering or a motion-robust readout | Replica moves with the encoding axis; position changes with TR (Figure 2); lessens with motion control |
| Anatomy from outside the image folded over the opposite edge | Aliasing (4.2) | Enlarge FOV, phase oversampling, swap phase-encoding direction | Disappears or moves; sharp rather than smeared |
| Parallel bright and dark bands beside a sharp high-contrast edge, spacing of a few pixels | Truncation (4.3) | Increase matrix; apply a filter; swap phase-encoding direction; repeat on a different sequence | Band spacing scales with pixel size (Figure 1); band absent in the other plane or sequence; may mimic syrinx in the cord [6] |
| Dark/bright rim at fat–water interface, one side bright and opposite side dark | Chemical shift, first kind | Change receiver bandwidth; swap frequency-encoding direction; apply fat suppression | Displacement in pixels ∝ 1/BWpx (Figure 3); rim moves or disappears |
| Signal void with distortion next to metal or air | Susceptibility (4.1) or metal [7] | Shorten TE; use spin echo instead of gradient echo; increase bandwidth; thinner slices; change orientation | Void size shrinks as predicted; true lesion would not |
| Thin line of alternating bright/dark pixels along the phase-encoding axis at a fixed position | RF interference (4.4) | Repeat with different anatomy or positioning; check door and equipment | Persists in the same image position regardless of anatomy; needs service investigation |
| Increased tendon signal at short TE, orientation about 55° to B0 | Magic angle (4.5) | Repeat at longer TE; reposition so the fibres leave 55° | Signal falls on long TE or after repositioning [3] |
| Vessel lumen signal that does not match expected flow | Inflow / outflow / dephasing | Add flow-compensation, change slice orientation, compare sequences | See 9023, 9207 |
| Broad intensity shading, centre bright and periphery dark at 3 T | B1 non-uniformity | Dielectric pads; change coil or sequence; apply intensity correction | Pattern stable between patients; changes with field strength (principle, 3.4) |
| Half-field-of-view ghost on echo-planar images | EPI Nyquist ghost | Check whether the ghost sits at N/2 along phase encoding independent of motion; compare with reference-scan correction on | Position fixed relative to the object; not moved by changing TR |
| Stretched, compressed or displaced anatomy along phase encoding on EPI/diffusion images | EPI susceptibility distortion | Reverse the phase-encoding direction; compare with a non-EPI sequence | Distortion reverses in direction; the non-EPI image is geometrically faithful |
| Abrupt sign reversal of velocity on a phase-contrast map | Velocity aliasing (VENC below peak velocity) | Repeat with higher VENC | Reversal disappears at the higher VENC |
| Periodic intensity bands across a multi-slab angiogram | Slab-boundary (venetian-blind) artefact | Change slab overlap or acquire as a single slab | Banding period follows slab spacing |
| Bright or dark tissue that is wrong for the sequence after a fat-suppression or inversion pulse | Suppression failure | Change method (e.g. inversion-based instead of spectral); re-shim; centre the anatomy | Pattern follows field uniformity, not anatomy |
| Geometric distortion growing with distance from isocentre | Gradient non-linearity [5] | Compare with the distortion-corrected reconstruction; centre the region of interest | Corrected series is geometrically accurate |
A finding that survives every one of these changes unchanged is more likely to be anatomy or pathology. A finding that moves with a predicted parameter is an artefact. One change is not enough when the finding is clinically decisive; the page for each artefact will say which change is the most discriminating.
8. Mitigation Levers and What Each Costs
No lever is free. The table gives the cost so that choosing a lever is a decision rather than a reflex; the costs are those that follow from the physics and from the parameter pages of the site (MRI Parameters).
| Lever | Reduces | Cost / limitation |
|---|---|---|
| Higher receiver bandwidth (9597) | Chemical-shift and susceptibility displacement in pixels and millimetres (Figure 3) | Lower SNR; may increase minimum echo spacing demands on gradients |
| Larger matrix at fixed FOV and bandwidth (9582) | Width of Gibbs rings (Figure 1) | Longer time or lower SNR per pixel; does not change displacement in mm at fixed total bandwidth (5.2) |
| Larger FOV or phase oversampling (9585, 9581) | Phase-direction aliasing [6] | Lower resolution at fixed matrix, or longer scan time |
| Swapping phase- and frequency-encoding directions (9593) | Moves ghosts, wrap and ringing away from a region of interest | Does not remove them; may bring aliasing into the other axis |
| Saturation (rest) slabs (9586) | Ghosts from flowing or moving tissue outside the region | SAR and time; may suppress wanted flow |
| Cardiac or respiratory triggering, gating or navigators | Periodic-motion ghosts | Longer and less predictable scan time; arrhythmia and irregular breathing reduce benefit |
| Radial or blade k-space (9028) | Motion sensitivity; ghosts become diffuse | Longer scan time; streak patterns replace ghosts; correction limited to modelled motion |
| Signal averaging (9589) | Intensity of non-periodic motion ghosts | Time; limited effect on bulk motion |
| Parallel imaging (9588) | Motion window and EPI distortion by shortening acquisition | g-factor noise and residual aliasing |
| Spin echo instead of gradient echo; shorter TE | Static susceptibility dephasing | Longer time, higher SAR (spin echo); less T2* contrast |
| Fat suppression | Chemical-shift displacement and fat ghosts | Sensitive to field and B1 non-uniformity; SNR cost; failure near metal [7] and wherever the field is inhomogeneous |
| Apodisation (filtering) of k-space | Gibbs ringing | Blurring: loss of sharpness at the edge |
| Patient preparation: positioning, coaching, immobilisation, removal of ferromagnetic items and make-up | Motion; foreign-body artefacts [6] | Staff time; not always feasible (children, acute illness) |
| Site engineering and quality assurance | Zipper and RF interference; coil failure; distortion; drift | Service call and downtime; requires defined escalation [4] |
Zaitsev and colleagues conclude for motion alone that a large toolbox of correction methods has accumulated over decades, none of which works in every situation [8]. The same is expected to hold for the other artefact families: the aim of the child pages is to say which lever fits which situation, not to claim a universal remedy.
9. Who Acts on What
Reasoning: a workable division follows from the controllability axis (Section 2.2). The technologist removes what is controllable at the console and from the patient; the radiologist decides whether a residual artefact limits the answer to the clinical question and states that in the report; the medical physicist or service engineer handles artefacts that persist across patients, protocols or coils. Zhuo and Gullapalli recommend regular testing and monitoring of equipment, testing of new or upgraded systems by a physicist or qualified engineer before clinical use, and participation in a recognised accreditation programme as the foundation of a quality-assurance programme [4]. A report should name an artefact when it limits the conclusion, state which structures are affected, and say what was done (or could be done) to resolve it.
10. MRI Technologist and Radiologist Pearls — Common Misconceptions
- “Susceptibility artefact and aliasing look similar, so they’re probably related phenomena.” They have entirely unrelated physical origins (Sections 4.1 and 4.2) — susceptibility artefact arises from genuine local field distortion at tissue interfaces, while aliasing arises from a k-space sampling-rate violation.
- “Truncation ringing means something is wrong with the scan.” It is a normal, expected, mathematically inevitable consequence of any finite k-space acquisition at a sharp, high-contrast boundary (Section 4.3) — not evidence of a technical fault. Its width can be reduced by a larger matrix, but its fractional overshoot cannot (Figure 1).
- “A zipper artefact is just another normal imaging artefact to expect on every scan.” Unlike susceptibility, aliasing, truncation ringing and the magic angle effect — all of which occur routinely even with well-maintained hardware — a genuine zipper artefact (Section 4.4) specifically indicates an RF shielding or interference problem that generally warrants site engineering investigation.
- “The magic angle effect can be eliminated by adjusting acquisition parameters like a typical artefact.” Because it arises from a genuine physical interaction dependent on anatomical geometry relative to B0, it is managed through recognition (echo-time dependence, characteristic distribution) rather than elimination (Section 4.5).
- “Increasing the matrix reduces the chemical-shift displacement.” At fixed field of view and fixed total receiver bandwidth the displacement in millimetres is unchanged (Section 5.2); it falls with higher bandwidth or removal of the offset.
- “Swapping phase and frequency directions fixes an artefact.” It relocates ghosts, wrap and ringing; it is a diagnostic manoeuvre and a way to move an artefact off a region of interest, not a correction.
- “A ghost at a given position tells you the patient moved at a given moment.” The ghost position depends on motion rate and TR together (Figure 2), not on anatomy; a different TR moves the ghost.
11. Advanced Technical Notes
11.1 Susceptibility Artefact Mitigation Strategies (retained)
Beyond sequence-family selection (spin-echo vs gradient-echo, Section 4.1.2), susceptibility-related signal loss and distortion can be further reduced through techniques including higher receiver bandwidth (reducing the spatial extent, in pixels, of a given frequency shift along the frequency-encode direction — the same bandwidth-vs-misregistration relationship introduced for chemical shift artefact in Chemical Shift and Fat/Water Physics), thinner slices, and specific susceptibility-artefact-reduction sequence variants — a topic connecting to Reconstruction and Post-Processing Physics for the more advanced correction methods. Hargreaves and colleagues state that careful choice of parameters and pulse sequences can avoid or reduce metal artefacts and that more advanced methods can improve results further [7].
11.2 Partial Fourier and Parallel Imaging’s Effect on Truncation Ringing (retained)
Because truncation ringing (Section 4.3) depends directly on how much of k-space’s high-spatial-frequency content is actually acquired, acceleration techniques that reduce the effectively acquired k-space extent — most directly, aggressive partial Fourier factors — can measurably worsen truncation ringing severity relative to a fully-sampled acquisition at the same nominal resolution, a practical interaction between two otherwise separately-discussed topics worth flagging explicitly.
11.3 Bandwidth, Displacement and SNR
Reasoning (derivation). For a fixed acquisition, noise variance in the image is proportional to the receiver bandwidth, so SNR falls as the square root of bandwidth, while the displacement in pixels falls in proportion to bandwidth. Halving the displacement therefore costs about 29% of the SNR (a factor of 1/√2) when bandwidth is doubled, before any change of echo time or gradient demands. Parameter pages of the site (9597) discuss the practical side; the derivation is given here because every off-resonance child page uses it.
Bibliography for this section
12. Roadmap of Child Pages
The master is the first step. Each planned page in Section 6 (60 pages for 64 table entries, because related entries share a page) will get its own child page, built to the same structure: mechanism, appearance, discriminating tests, mitigation and cost, protocols or anatomical regions where it matters, evidence and limits, references. Child pages will assume this page and will not repeat Sections 3–8. IDs 9601–9699 are reserved provisionally; they become final when a page is published. Priority tiers combine clinical frequency, risk of misdiagnosis and how much of the mechanism is not yet on the site; the tiering is a proposal for discussion, not a published ranking.
Tier 1 — highest yield (proposed first)
| Proposed ID | Child page | Scope beyond this master | Link now |
|---|---|---|---|
| 9601 | Bulk voluntary and involuntary motion | Quantitative ghost mechanics beyond Figure 2; tolerance by readout; link to blade methods | 9207 · 9028 |
| 9602 | Respiratory motion | Chest, abdomen, spine; breath-hold, gating, navigators, radial | 9207 · 9586 |
| 9603 | Cardiac and arterial pulsation | Vessel-origin ghosts; flow compensation; triggering | 9207 |
| 9604 | CSF pulsation | Spine and posterior fossa; T2 and FLAIR | — |
| 9605 | Metal implant artefacts | Sequence families for metal; reporting | 9004 |
| 9606 | Susceptibility (air–bone, blood products, calcification) | Sequence dependence and quantification | 9011 |
| 9607 (published) | Chemical shift, first kind (misregistration) | Bandwidth, field-strength and orientation dependence | 9206 · 9596 · 9597 |
| 9608 (published) | Chemical shift, second kind (India-ink / opposed phase) | Opposed-phase cancellation and echo-time tables | 9206 |
| 9609 | Fat-suppression failure | Causes by method; remedies | 9006 |
| 9610 | Truncation (Gibbs) ringing; syrinx pitfall | Spine and musculoskeletal; filters; partial Fourier interaction | Section 4.3 |
| 9611 | Aliasing / wrap-around (phase, frequency and slice directions) | Recognition, 3D slab wrap, parallel-imaging aliasing | 9585 · 9594 · 9593 · 9595 |
| 9612 | EPI Nyquist (N/2) ghost | Mechanism, correction, residual | 9010 |
| 9613 | EPI geometric distortion and fat displacement | Susceptibility, reversed phase-encoding correction | 9010 · 9014 |
| 9614 | Eddy-current artefacts | Diffusion gradients; correction | 9208 |
| 9615 | Parallel-imaging noise (g-factor) and residual aliasing | Noise amplification maps; failure modes | 9588 |
Tier 2 — important, narrower
| Proposed ID | Child page | Scope beyond this master | Link now |
|---|---|---|---|
| 9616 | Slow flow, inflow and entry-slice effects; T1 shine-through | Differentiation from thrombus and occlusion | 9023 · 9213 |
| 9617 | Flow voids, intravoxel dephasing and flow misregistration | Stenosis versus pseudo-occlusion | 9207 |
| 9618 | B1 inhomogeneity and dielectric/standing-wave shading | 3 T body and head | 9211 · 9208 |
| 9619 | Receive-coil intensity bias, element drop-out and loading errors | Coil faults and normalisation | — |
| 9620 | Gradient non-linearity distortion | Corrections; measurement implications | 9208 |
| 9621 | RF interference: zipper and k-space spikes | Troubleshooting algorithm | Section 4.4 |
| 9622 | Magic angle effect | Tendons, ligaments, menisci, nerves | Section 4.5 |
| 9623 | Slice cross-talk, cross-excitation and slice-profile effects | Gap and interleaving | 9591 · 9583 |
| 9624 | Blooming and phase wrap on gradient echo, SWI and QSM | Quantitative TE dependence | 9011 · 9016 |
| 9625 | bSSFP banding | Phase cycling and field dependence | 9015 |
| 9626 | FLAIR-specific artefacts (CSF flow, supplemental oxygen, vessels) | CSF flow, supplemental oxygen, vessels | 9007 |
| 9627 | T2 shine-through and T2 blackout | Interpretive pitfalls | 9012 · 9013 |
| 9638 | Swallowing, tongue and neck motion | Head and neck; cervical spine | — |
| 9639 | Eye movement and blinking | Brain and orbit; phase-direction choice | — |
| 9641 | Pulsation and motion signal loss in diffusion-weighted imaging | Cardiac-induced signal loss in body and brain DWI | 9012 · 9013 |
| 9642 | Transient severe motion after gadoxetate | Liver MRI with hepatobiliary agent; evidence appraisal | 9101 |
| 9643 | Phase-contrast velocity aliasing and background phase errors | Velocity wrap and phase-offset correction | 9022 |
| 9644 | Slab-boundary (venetian-blind) artefact in 3D multi-slab TOF | MOTSA and slab design | 9024 · 9025 · 9599 |
| 9645 | Contrast-enhanced MRA timing artefacts | Timing and k-space ordering | 9101 |
| 9649 | Radial, spiral and blade undersampling streaks | Streak mechanisms and mitigation | 9028 |
| 9650 | Stitching artefacts in multistation imaging | Whole-spine and whole-body imaging | — |
| 9651 | Partial-volume averaging | Pitfall rather than defect: quantitative treatment | 9583 |
| 9654 | Incomplete nulling in STIR and FLAIR | Inversion-recovery failure modes | 9006 · 9007 · 9020 |
| 9655 | Dental and orthodontic hardware | Head and neck and brain | — |
| 9658 | Poor shimming and off-centre imaging | Shim volume and field-of-view placement | 9208 |
| 9659 | Turbo spin-echo blurring and bright fat | Echo-train effects | 9005 |
Tier 3 — specialised or evidence still developing
| Proposed ID | Child page | Scope beyond this master | Link now |
|---|---|---|---|
| 9628 | Subtraction misregistration | Dynamic contrast-enhanced studies | 9101 |
| 9629 | Partial Fourier, zero-filling, interpolation and filter effects | What interpolation can and cannot do | 9587 |
| 9630 | Compressed-sensing and deep-learning reconstruction artefacts | Appraisal of an emerging and heterogeneous literature | 9212 |
| 9631 | Moiré (fold-over) fringes in gradient echo | Gradient echo with aliasing | — |
| 9632 | Dixon fat–water swap | Mechanism and detection | — |
| 9633 | Foreign bodies and patient-related sources | Foreign bodies in practice | — |
| 9634 | Vibration, B0 drift and table motion | Stability faults | 9208 |
| 9635 | Quality assurance for artefact detection | Phantoms, routine checks, escalation | — |
| 9636 | Cardiac-MR artefact set | Gating, arrhythmia, banding | — |
| 9637 | Peristalsis and bowel gas | Abdomen and pelvis | 9586 |
| 9640 | Fetal and neonatal motion | Fetal and neonatal imaging | — |
| 9646 | Arterial spin labelling artefacts | ASL pitfalls | 9021 · 9301 |
| 9647 | Central point (zero-frequency) artefact | Hardware origin | — |
| 9648 | Receiver overload and data clipping | Gain setting and clipping | — |
| 9652 | Reformat and projection artefacts (MIP, MPR, volume rendering) | MIP covered; MPR and rendering to be added | 9027 |
| 9653 | Chemical shift in the slice-selection direction | Slice-direction misregistration | — |
| 9656 | Concentrated gadolinium T2* effects | Excretory phase pitfall | 9101 |
| 9657 | Diffusion b-value and ADC bias from gradient non-linearity | Quantitative diffusion | 9208 |
| 9660 | MR spectroscopy artefacts | Spectroscopy pitfalls | — |
Each child page will be researched from primary sources and checked at abstract level at least before citation; items for which only manufacturer or conference material exists will say so. Pages for metal implants, EPI and chemical shift are likely to have the broadest literature; pages on deep-learning reconstruction artefacts are likely to have the weakest and least settled.
13. Evidence Gaps and Ongoing Debate
- No society guideline or consensus statement defines or classifies MRI artefacts. The content of this page rests on physics, technical reviews and educational articles [1–8], which are low in the clinical evidence hierarchy but appropriate for mechanism; clinical statements about how often an artefact leads to a wrong diagnosis are not supported by data retrieved for this page.
- No prevalence or cost data were found for the artefacts as a group. Studies measuring the frequency, repeat rate or cost of individual artefacts exist for some contexts but were not retrieved here and are left to the child pages.
- Optimal susceptibility and metal-artefact correction remains an active research area, with ongoing development for high-field and near-metal imaging rather than a single settled, universally adopted approach.
- Quantitative characterisation of the magic angle effect’s clinical significance across tissue types is less thoroughly characterised outside the classic tendon and ligament literature.
- Zipper and RF-interference source identification remains institution-dependent troubleshooting rather than a standardised physics-based diagnostic algorithm.
- Deep-learning and compressed-sensing reconstruction introduce a class of artefacts that is not yet well described; the evidence is expected to be heterogeneous and vendor-specific.
- Models in Figures 1–3 are idealised. They give scaling relations (ring spacing, ghost position, displacement) and not the amplitude of a given artefact on a given scan.
- Retrieval limits in this revision. Sources [4]–[8] were read at abstract or retrievable-text level; full text was not available for all. The volume and page numbers of [7] were not shown on the source page and are not stated.
14. Evidence-Based References
A. Guidelines / Consensus / Society Recommendations
No society guideline or consensus statement on the classification of MRI artefacts was identified. Category A is therefore not populated.
B. Systematic Reviews and Meta-Analyses
None retrieved for artefact physics as a topic in this revision. Category B is not populated.
C. Technical MRI Papers and Educational Reviews (narrative)
D. Foundational and Historical Landmarks
End of document — MRI Artefacts (Master) — MRIninja v2.1 — October 2026. Supersedes v1.0 “Physics of MRI Artefacts” (July 2026).
Child Protocols
Clinical pages derived from this master protocol. These pages document what changes for specific indications.
Related Protocols
Recent PubMed search for this protocol