MRI Image Artifacts: Identification & QC

MRI image artifacts are systematic signal errors introduced by the scanner, the pulse sequence, or the patient—not by anatomy—and telling them apart from disease is a core quality-control task. ^{1, 2} A defensible MRI quality program recognizes each artifact by its mechanism, separates an expected physics effect from a genuine hardware fault, and documents the phantom measurements that prove the system is performing within tolerance.

Introduction

Almost every MRI examination contains at least one artifact. ³ Some are harmless; others mimic pathology, obscure the finding the study was ordered to answer, or signal that the scanner itself is drifting out of specification. The clinical and safety stakes are real: a truncation artifact can be read as a spinal cord syrinx, a susceptibility artifact from a forgotten hairpin can hide an acute infarct, and a slow decline in signal-to-noise ratio can erode diagnostic confidence long before a radiologist notices it. ^{3, 10}

For the medical physicist and the MRI quality-control program, the essential skill is not memorizing a catalog of picture patterns—it is understanding the physics that produces each artifact. Once you know that ghosting propagates along the phase-encoding direction, that chemical shift scales with field strength, or that metal distorts the local magnetic field, you can predict where an artifact will appear, decide whether it is expected or abnormal, and choose the acquisition change that reduces it. ^{1, 2}

This guide walks through the major MRI artifact families, the technical principles behind them, and the American College of Radiology (ACR) phantom quality-control metrics that keep an accreditation program defensible. DRPS provides annual MRI performance evaluations and QC program support as part of its MRI physics testing and medical physicist consulting services across Florida, Maryland, Virginia, Washington DC, California, and Nevada.

Topic Explanation

What is an MRI artifact?

An MRI artifact is any feature in the image that does not correspond to the true spatial distribution of tissue. ^{1, 3} It is the difference between what the scanner reconstructed and what is physically present in the patient. Artifacts arise wherever an assumption built into the imaging process breaks down—when the patient moves during acquisition, when tissue is not where the frequency map expects it, when the magnetic field is distorted by metal, or when the receiver samples too coarsely to represent a sharp edge.

Artifacts fall into three broad origin groups, and sorting a finding into the right group is the fastest route to a fix: ^{1, 2, 3}

Patient-related: motion (breathing, cardiac, vascular pulsation, swallowing, gross movement), metallic foreign bodies and implants, and inherent tissue interfaces such as fat and water or air and soft tissue.
Sequence- and sampling-related: aliasing (wraparound), Gibbs truncation, and chemical shift—effects that come from how k-space is sampled and how spatial position is encoded.
Hardware- and system-related: gradient nonlinearity, radiofrequency (RF) interference, coil element failure, and main-field instability.

The distinction matters operationally. A patient-related artifact is managed with positioning, screening, and sequence choice; a sequence-related artifact is managed by changing acquisition parameters; a hardware-related artifact is a QC finding that may require service. For the physics of how the underlying image is formed—and why some artifacts are unavoidable—see our companion discussions of MRI geometric distortion QC and MRI SNR and RF coil QC.

Why artifact recognition is a safety issue, not just an aesthetic one

The danger of an artifact is not that it looks bad—it is that it can be misread. ^{3, 7, 8} A chemical-shift band can simulate a cortical rim; a pulsation ghost can project over a vessel and mimic dissection; a metal susceptibility artifact can erase the very region a study was ordered to evaluate. The reader who does not recognize the artifact may act on it, and the reader who over-attributes a real finding to "just an artifact" may dismiss disease. Both errors are prevented by the same thing: a shared, physics-based vocabulary for what each artifact is and where it belongs.

Key Technical Principles

MRI encodes spatial position with magnetic field gradients: frequency along one axis, phase along another. Almost every artifact is a breakdown of that encoding. The table below maps the major artifact families to their mechanism, appearance, and the primary lever the physicist or technologist uses to reduce them. ^{1, 2, 3, 6}

Artifact	Physical mechanism	Typical appearance	Primary mitigation
Motion / ghosting	Patient or physiologic motion between phase-encoding steps	Discrete replicas ("ghosts") spread along the phase-encoding direction	Immobilization, cardiac/respiratory gating, saturation bands, faster sequences (EPI), swap phase/frequency direction ³
Chemical shift	~3.5 ppm precession-frequency difference between fat and water	Dark/bright band at fat–water interfaces along frequency-encoding	Wider receiver bandwidth, lower field, fat suppression, smaller voxel ⁵
Susceptibility / metal	Local field distortion near metal or air–tissue interfaces	Signal void with bright pile-up rim and geometric warping	Spin-echo with short TE, higher bandwidth, avoid gradient echo/EPI, metal-suppression sequences (STIR, VAT, SEMAC, MAVRIC) ⁶
Gibbs / truncation	Finite k-space sampling near a high-contrast edge	Parallel alternating bright/dark lines ("ringing") near sharp boundaries	Increase matrix (phase-encoding steps), apply k-space filtering ^{3, 12}
Aliasing / wraparound	Anatomy outside the field of view folds to the opposite side	Structure from one edge superimposed on the other	Increase FOV, oversample, use no-phase-wrap options ³
RF / zipper and spike	External RF leakage or gradient/static-electricity spikes	Bands of noise ("zipper") or herringbone/checkerboard pattern	Confirm RF shield/door seal integrity, investigate hardware; screen clothing/static ^{3, 8}
Gradient nonlinearity	Gradient field departs from ideal linearity toward the FOV edge	Geometric distortion, worse peripherally and at higher field	Vendor distortion correction, keep anatomy near isocenter ¹

Chemical shift: a worked example

Fat and water protons resonate at slightly different frequencies because their electron environments shield the nucleus differently; the separation is approximately 3.5 parts per million (ppm). ⁵ Because MRI assigns spatial position by frequency along the readout axis, the scanner places fat a fixed distance away from its true location. The frequency difference is:

Δ f_{CS} = δ \cdot (\frac{γ}{2 π}) \cdot B_{0}

where $δ \approx 3.5 \times 1 0^{- 6}$ is the chemical-shift difference, $γ /2 π \approx 42.58 MHz/T$ is the proton gyromagnetic ratio, and $B_{0}$ is the field strength. ^{5, 12} At 3 T:

Δ f_{CS} = (3.5 \times 1 0^{- 6}) (42.58 \times 1 0^{6} Hz/T) (3 T) \approx 447 Hz

The spatial displacement in pixels is that frequency difference divided by the receiver bandwidth per pixel:

d_{CS} = \frac{Δ f _{CS}}{BW _{pixel}}

With a receiver bandwidth of 200 Hz/pixel, fat is displaced by $447/200 \approx 2.2$ pixels at 3 T. At 1.5 T the same calculation gives about 224 Hz and roughly half the displacement. Two levers follow directly from the equation: chemical shift is worse at higher field, and it shrinks when you widen the receiver bandwidth—the standard fix when a chemical-shift band obscures a fat–water boundary. ⁵

Percent signal ghosting: the QC metric

Ghosting is the artifact the ACR phantom program quantifies directly. On the ACR MRI accreditation phantom—a 190 mm inside-diameter, 148 mm-long acrylic cylinder filled with a nickel-chloride and sodium-chloride solution—the percent signal ghosting (PSG) test places one large signal region of interest (ROI) at the center of the uniform slice and four background ROIs above, below, left, and right of the phantom. ¹⁰ PSG is computed as:

PSG = \frac{∣ ( S _{top} + S _{bottom} ) - ( S _{left} + S _{right} ) ∣}{2 S _{ROI}}

where each $S$ is a mean ROI signal. ¹⁰ To pass, the ghosting ratio must be ≤ 2.5% on the ACR phantom; many acceptance-testing protocols apply a stricter ≤ 1% limit. ¹⁰ Rising PSG points to system instability—gradient or RF imperfections, or eddy-current problems—and is one of the earliest quantitative signs that a scanner is drifting.

The ACR image-quality battery pairs PSG with percent image uniformity (PIU), which must be ≥ 87.5% for systems below 3 T (AAPM protocols use ≥ 90%), and with signal-to-noise measurements. ¹⁰ SNR is not part of the ACR test set itself, but it is central to acceptance testing; the NEMA single-image method multiplies the ratio of mean phantom signal to background standard deviation by 0.655—a factor that corrects for the Rayleigh distribution of background noise in magnitude images—with a common pass criterion of SNR ≥ 80 × $B_{0}$ (in tesla). ¹⁰

Clinical Impact

Artifact literacy changes reads at the point of care. In the orbit, chemical shift at the fat–globe interface and motion from eye movement are expected; recognizing them prevents over-calling a lesion and guides fat suppression when a true mass is suspected. ⁷ In the spine, truncation ringing parallel to the cord can imitate a syrinx, and CSF pulsation ghosts can mimic an intradural lesion. ^{3, 8} In the breast, wraparound, motion misregistration, and failed fat suppression are the dominant pitfalls, and they directly affect whether an enhancing focus is judged real. ⁹ Around arthroplasty and other metal implants, susceptibility artifact can obliterate the periprosthetic tissue where infection or loosening lives—so metal-suppression sequences are not a luxury but the difference between a diagnostic and a non-diagnostic study. ⁶

Some artifacts are even turned to diagnostic advantage. The chemical-shift signal drop on opposed-phase gradient-echo imaging confirms microscopic fat and is the basis for characterizing adrenal adenomas; the same physics that creates a nuisance band at one interface confirms a diagnosis at another. ^{3, 5} The physicist who understands the mechanism can explain to the reading radiologist not only how to suppress an artifact but when to exploit it.

At the program level, artifacts are the visible tip of image-quality drift. A scanner that begins to ghost, lose uniformity, or show intermittent zipper lines is telling the QC program something about its gradients, RF chain, or shielding before those problems reach every clinical image. That is why artifact assessment is embedded in phantom QC rather than left to case-by-case interpretation. ^{2, 10}

Practical Optimization Tips

A practical MRI artifact and image-quality workflow follows a consistent sequence.

1. Localize the artifact to a direction

Determine whether the artifact runs along the phase-encoding or frequency-encoding axis. Ghosting propagates along phase encoding; chemical shift and most bandwidth-dependent effects sit along frequency encoding. This single observation narrows the cause immediately. ^{1, 3}

2. Test whether it moves with parameters

Swap the phase- and frequency-encoding directions, change the field of view, or change the receiver bandwidth. An artifact that moves or scales with these changes is sequence-related; one that stays fixed relative to hardware is a system finding. ^{1, 3}

3. Match the sequence to the hazard

For motion, add gating or saturation bands, shorten the acquisition, or move the phase-encoding direction so ghosts project away from the region of interest. ³
For chemical shift, widen the receiver bandwidth and consider fat suppression. ⁵
For metal, use spin-echo with a short echo time and high bandwidth, avoid gradient-echo and echo-planar sequences, and deploy metal-suppression techniques (STIR, VAT, SEMAC, MAVRIC). ⁶
For truncation ringing, increase the phase-encoding matrix. ^{3, 12}
For wraparound, increase the FOV or turn on no-phase-wrap oversampling. ³

4. Screen the patient and the room

Many "artifacts" are avoidable: metallic thread in clothing, cosmetics containing iron oxide, hair accessories, and static-generating fabrics all produce signal loss or spike noise. ³ A disciplined screening and pocketless-attire policy removes a large fraction of susceptibility and spike artifacts before the scan starts—an operational point that ties artifact control directly to the MRI safety program.

Common pitfalls to avoid

Reading an artifact as pathology (or vice versa). Always test whether a suspicious finding follows the physics of the acquisition before attributing it to disease. ^{3, 7}
Ignoring a rising QC trend. A PSG or uniformity value still inside tolerance but trending in one direction is an early warning, not a pass to forget. ¹⁰
Using gradient-echo near metal. Gradient-echo and EPI accentuate susceptibility; they are the wrong tool around implants. ⁶
Assuming higher field is always better. Chemical shift, susceptibility, and geometric distortion all worsen at 3 T; the QC program must account for field strength. ^{1, 5}
Skipping the physicist's interpretation. Phantom numbers without expert interpretation miss the difference between an expected physics effect and an incipient hardware fault. ^{2, 10}

Regulatory Considerations

MRI does not have an MQSA-style federal quality mandate, so the binding image-quality requirements come from accreditation and hospital-standards bodies rather than a single X-ray regulation. ² Unlike CT or radiography, MRI uses no ionizing radiation, so it falls outside the NRC's byproduct-material rules and outside the X-ray machine registration programs administered by the FDA and the states. Its quality framework is built instead on ACR accreditation and Joint Commission standards.

Key frameworks to reference:

ACR MRI Accreditation Program and the ACR MRI Quality Control Manual — define the ACR phantom test battery (geometric accuracy, high-contrast resolution, slice thickness and position accuracy, image uniformity, percent signal ghosting, low-contrast detectability, and a visual artifact check) and require an annual performance evaluation by a qualified medical physicist. ¹⁰
AAPM Report No. 100 — the profession's acceptance-testing and QA reference for MRI facilities, used alongside the ACR phantom for commissioning and troubleshooting. ¹⁰
NEMA measurement standards — define reproducible methods for SNR and image uniformity that underpin acceptance testing. ¹⁰
Joint Commission imaging standards — require an annual MRI equipment performance evaluation by a qualified medical physicist or MR scientist for accredited organizations.

Because the states DRPS serves regulate ionizing-radiation machines but not MRI's magnetic and RF fields, an imaging center's MRI image-quality obligations are driven by its accreditation status and hospital contracts. A documented QC program—weekly technologist phantom checks, an annual physicist evaluation, and a record of artifact findings and corrective actions—is what makes the program defensible during an accreditation review. For how the same accreditation logic applies across modalities, see ACR accreditation physics requirements.

Frequently Asked Questions (FAQs)

What causes most MRI image artifacts?

Most MRI artifacts come from one of three sources: the patient (motion, metal implants, tissue interfaces), the pulse sequence and its sampling (aliasing, Gibbs truncation, chemical shift), or the scanner hardware (gradient nonlinearity, RF interference, coil failure). Identifying which of the three is responsible is the first step in correcting the artifact or deciding it is an expected physics effect rather than a fault.

How is an MRI artifact different from a real lesion?

Artifacts usually follow the physics of the acquisition rather than anatomy: they align with the frequency- or phase-encoding direction, repeat at regular intervals, appear only on certain sequences, or move when acquisition parameters change. A finding that shifts with the frequency-encoding direction, disappears on a spin-echo sequence, or tracks a metal implant is almost always an artifact, not pathology.

What is percent signal ghosting on the ACR MRI phantom?

Percent signal ghosting (PSG) measures faint replicas of the phantom that appear in the background along the phase-encoding direction, usually from system instability. It is calculated from four background regions of interest relative to the central phantom signal. On the ACR accreditation phantom the action limit is 2.5% or less; many acceptance-testing protocols apply a tighter 1% limit.

Why do fat and water separate at tissue boundaries on MRI?

Fat and water protons precess at slightly different frequencies—about 3.5 parts per million apart—because of their different chemical environments. The scanner assigns spatial position by frequency, so it misregisters fat relative to water along the frequency-encoding direction, producing a dark or bright band at fat–water interfaces. The displacement grows with field strength and with narrower receiver bandwidth.

Can MRI artifacts be eliminated completely?

No. Some artifacts, such as chemical shift and susceptibility, are direct consequences of MR physics and can only be reduced, not removed. The goal is to minimize artifacts that obscure anatomy, recognize the ones that remain, and use them diagnostically when possible—for example, using chemical-shift signal loss to confirm microscopic fat in an adrenal adenoma.

How often should MRI artifact and image-quality QC be performed?

Under the ACR MRI accreditation program, a qualified medical physicist performs a full annual performance evaluation, and technologists run weekly ACR phantom QC that includes a visual check for artifacts. New systems, coil replacements, software upgrades, and any change in image quality should trigger additional testing outside the routine schedule.

Does a medical physicist need to be involved in MRI QC?

Yes. The ACR MRI accreditation program and Joint Commission imaging standards require an annual MRI equipment performance evaluation by a qualified medical physicist or MR scientist. The physicist interprets the phantom measurements, distinguishes hardware faults from physics-based artifacts, and directs corrective action before image quality affects diagnosis.

Key Takeaways

Artifacts are systematic, not random. Each follows a physical mechanism, so recognizing the mechanism predicts the appearance and points to the fix. ^{1, 2}
Direction is diagnostic. Ghosting runs along phase encoding; chemical shift runs along frequency encoding. Localizing the axis narrows the cause immediately. ^{1, 3}
Chemical shift and susceptibility worsen at higher field. They can be reduced with bandwidth, sequence choice, and suppression—but not eliminated. ^{5, 6}
Percent signal ghosting is a quantitative early-warning metric. The ACR phantom limit is ≤ 2.5% (≤ 1% in many acceptance protocols); a rising trend flags system instability. ¹⁰
Metal calls for spin-echo and suppression, not gradient echo. Gradient-echo and EPI accentuate susceptibility artifact. ⁶
MRI QC is accreditation-driven. With no MQSA equivalent, ACR accreditation and Joint Commission standards—anchored by an annual physicist evaluation—define the requirements. ¹⁰

Conclusion

MRI artifacts are not a nuisance to be memorized as pictures; they are the visible signature of the physics that forms the image and of the hardware that produces it. The physicist and technologist who can name the mechanism behind ghosting, chemical shift, susceptibility, truncation, aliasing, and zipper artifacts can do three things a checklist cannot: separate an expected effect from a genuine fault, choose the acquisition change that reduces the artifact, and catch a scanner drifting out of specification before it reaches the reading room.

That understanding is what turns a phantom QC program from a compliance exercise into a real safeguard for diagnostic quality. A defensible MRI quality program pairs weekly phantom checks and an annual physicist evaluation with a documented record of artifact findings and corrective actions—so that when image quality is questioned, the answer is measured, not guessed.

How DRPS Can Help

Diagnostic Radiation Physics Services helps imaging facilities build MRI quality programs that are consistent, documented, and accreditation-ready. This includes annual MRI equipment performance evaluations, ACR phantom QC setup and training, artifact troubleshooting, acceptance testing of new systems and coils, and support for ACR accreditation submissions—delivered through our MRI physics testing, accreditation support, and medical physicist consulting services.

DRPS supports facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, Nevada, New York, Pennsylvania, New Jersey, and Delaware.

A strong MRI QC program does not just pass accreditation—it protects every read by making image quality measurable and every artifact explainable.

Related Resources

References

Morelli JN, Runge VM, Ai F, Attenberger U, Vu L, Schmeets SH, Nitz WR, Kirsch JE. An image-based approach to understanding the physics of MR artifacts. RadioGraphics. 2011;31(3):849-866. doi:10.1148/rg.313105115. PubMed
Zhuo J, Gullapalli RP. AAPM/RSNA physics tutorial for residents: MR artifacts, safety, and quality control. RadioGraphics. 2006;26(1):275-297. doi:10.1148/rg.261055134. PubMed
Krupa K, Bekiesińska-Figatowska M. Artifacts in magnetic resonance imaging. Polish Journal of Radiology. 2015;80:93-106. doi:10.12659/PJR.892628. PubMed
Pusey E, Lufkin RB, Brown RK, Solomon MA, Stark DD, Tarr RW, Hanafee WN. Magnetic resonance imaging artifacts: mechanism and clinical significance. RadioGraphics. 1986;6(5):891-911. doi:10.1148/radiographics.6.5.3685515. PubMed
Bley TA, Wieben O, François CJ, Brittain JH, Reeder SB. Fat and water magnetic resonance imaging. Journal of Magnetic Resonance Imaging. 2010;31(1):4-18. doi:10.1002/jmri.21895. PubMed
Talbot BS, Weinberg EP. MR imaging with metal-suppression sequences for evaluation of total joint arthroplasty. RadioGraphics. 2016;36(1):209-225. doi:10.1148/rg.2016150075. PubMed
Herrick RC, Hayman LA, Taber KH, Diaz-Marchan PJ, Kuo MD. Artifacts and pitfalls in MR imaging of the orbit: a clinical review. RadioGraphics. 1997;17(3):707-724. doi:10.1148/radiographics.17.3.9153707. PubMed
Taber KH, Herrick RC, Weathers SW, Kumar AJ, Schomer DF, Hayman LA. Pitfalls and artifacts encountered in clinical MR imaging of the spine. RadioGraphics. 1998;18(6):1499-1521. doi:10.1148/radiographics.18.6.9821197. PubMed
Ojeda-Fournier H, Choe KA, Mahoney MC. Recognizing and interpreting artifacts and pitfalls in MR imaging of the breast. RadioGraphics. 2007;27(Suppl 1):S147-S164. doi:10.1148/rg.27si075516. PubMed
Epistatou AC, Tsalafoutas IA, Delibasis KK. An automated method for quality control in MRI systems: methods and considerations. Journal of Imaging. 2020;6(10):111. doi:10.3390/jimaging6100111. PubMed
Sundermann B, Billebaut B, Bauer J, et al. Practical aspects of novel MRI techniques in neuroradiology: Part 1—3D acquisitions, Dixon techniques and artefact reduction. RöFo. 2022;194(10):1100-1108. doi:10.1055/a-1800-8692. PubMed
Currie S, Hoggard N, Craven IJ, Hadjivassiliou M, Wilkinson ID. Understanding MRI: basic MR physics for physicians. Postgraduate Medical Journal. 2013;89(1050):209-223. doi:10.1136/postgradmedj-2012-131342. PubMed