MRI Gradient and Eddy-Current QC

Introduction

In MRI, the gradient system is what turns a uniform magnetic field into an image, so any error in gradient calibration, linearity, or eddy-current compensation is written directly into the geometry and the quantitative values of every scan. Gradient quality control is the part of an MRI QC program that keeps that encoding honest.

Most MRI QC conversations start with signal-to-noise ratio, uniformity, and the familiar American College of Radiology (ACR) phantom score sheet. Those matter. But underneath every one of those measurements is a gradient subsystem quietly assuming that a commanded field of, say, 30 mT/m is exactly 30 mT/m, that the field varies linearly across the whole imaging volume, and that when the gradient is switched off it actually goes to zero without a lingering eddy-current tail. When any of those assumptions drifts, the scanner does not throw an error. It just produces images that are subtly the wrong size, warped at the edges, or carrying biased apparent diffusion coefficient (ADC) values. ¹³

This guide explains what the gradient system does, how gradient calibration, gradient nonlinearity, and eddy currents each fail, which QC tests detect each failure, and why the ACR phantom — while essential — is only part of a complete gradient QC picture. DRPS performs this evaluation as part of its MRI physics testing and accreditation support services across Florida, Maryland, Virginia, Washington DC, California, Nevada, Pennsylvania, New York, New Jersey, and Delaware.

Topic Explanation

What the gradient system actually does

MRI localizes signal by superimposing controlled, spatially varying magnetic fields — gradients — on top of the main field. Three orthogonal gradient coils (Gx, Gy, Gz) each produce a field whose z-component changes approximately linearly with position. That linear variation makes resonant frequency and phase depend on location, which is how slice selection, frequency encoding, and phase encoding work. Clinical whole-body gradient systems typically provide maximum amplitudes on the order of 30–80 mT/m and slew rates of roughly 100–200 T/m/s. ³

The encoding only works if the relationship between commanded gradient and physical field is known and stable. Three things can break that relationship:

Gradient calibration — the scaling factor between the digital gradient request and the actual field it produces. If this scale is off, distances in the image are systematically wrong.
Gradient linearity — how faithfully the field stays proportional to position across the whole field of view. Real gradient coils are most linear near isocenter and progressively nonlinear toward the periphery.
Eddy currents — currents induced in conductive structures (cryostat, shields, the coil former) whenever the gradient switches. They create their own transient fields that add to and lag behind the intended gradient.

For the tests that verify overall spatial fidelity, see our companion guide on MRI geometric distortion QC; this article focuses on the gradient behaviors that drive that distortion.

Why gradient QC is a quantitative-imaging problem, not just a geometry problem

It is tempting to treat gradient QC as purely about whether anatomy looks the right shape. That undersells it. Because the diffusion b-value scales with the square of gradient amplitude, gradient errors propagate into every gradient-weighted measurement — most importantly the ADC used in oncology, stroke, and treatment-response imaging. Published gradient-calibration work has shown that a mis-scaled gradient system produces ADC errors of nearly ±9% that are correctable to about ±1% once calibration is restored. ⁶ The gradient subsystem is therefore a foundational element of quantitative MRI QC, not just a cosmetic one.

Key Technical Principles

Gradient calibration and geometric scaling

If a gradient coil is commanded to produce amplitude $G_{0}$ but actually produces $G = G_{0} (1 + ε)$ , then the reconstruction — which assumes $G_{0}$ — maps physical position $x$ to an apparent position that is scaled by $1/ (1 + ε)$ . To first order, the fractional error in a measured length along that axis is:

\frac{Δ L}{L} \approx - ε = - \frac{Δ G}{G}

So a gradient that is 2% too strong compresses measured distances by about 2%. This is exactly what the ACR geometric accuracy test probes: the technologist measures known internal lengths of the phantom and compares them to the true values, with an action level near ±2 mm. ¹

The quantitative stakes are higher in diffusion imaging. For a standard pulsed-gradient spin-echo diffusion preparation, the b-value is:

b = γ^{2} G^{2} δ^{2} (Δ - \frac{δ}{3})

where $γ$ is the gyromagnetic ratio, $G$ the diffusion gradient amplitude, $δ$ the gradient duration, and $Δ$ the separation between the two gradient lobes. Because $b \propto G^{2}$ , a gradient scaling error $ε$ changes the b-value by approximately $2 ε$ . The measured ADC follows from $ADC = - ln (S / S_{0}) / b$ , so:

\frac{Δ ADC}{ADC} \approx - \frac{Δ b}{b} \approx - 2 ε

A worked example. Suppose a service event leaves the diffusion gradient miscalibrated by $ε = + 5%$ . The b-value is then high by about $2 \times 5% = 10%$ , and every ADC value is reported roughly 10% low. A true white-matter ADC of $0.80 \times 1 0^{- 3} m m^{2} /s$ would be reported near $0.72 \times 1 0^{- 3} m m^{2} /s$ — enough to move a lesion across a diagnostic threshold. Gradient-calibration measurements based on diffusion phantoms confirm this magnitude: scaling offsets of up to 5% produced ADC errors from −9.2% to +8.8%, reduced to within about ±1% after calibration. ⁶

Gradient nonlinearity

No gradient coil is perfectly linear. The field is engineered to be most linear within a specified diameter-of-spherical-volume (DSV) around isocenter and to fall away from ideal toward the bore edges. Uncorrected, this warps anatomy outward or inward at the periphery and biases quantitative values off-center. Vendors ship spatial-distortion (gradient nonlinearity) correction that applies a known field map to unwarp images, but the correction is only as good as the model and must actually be enabled.

The off-isocenter penalty is measurable. A diffusion-phantom study across positions found that applying gradient nonlinearity correction reduced ADC bias, with the largest absolute corrections at off-center positions and greater than 5% ADC bias observed at the extremes when correction was inadequate. ⁷ Distortion-correction studies at high field similarly show that gradient-nonlinearity unwarping should be applied whenever regions of interest sit well away from isocenter. ⁸ The practical lesson: geometric and quantitative accuracy validated at isocenter does not automatically extend to the corners of the field of view.

Eddy currents and their compensation

When a gradient switches on or off, the changing flux induces eddy currents in surrounding conductors. Those currents generate transient magnetic fields that oppose the change, so the effective gradient rises and falls more slowly than commanded and carries a decaying tail after the pulse ends. Left uncompensated, eddy currents cause:

Image shear, scaling, and shift in EPI-based diffusion, because the eddy-current field differs for +/− diffusion gradient polarities and across b-values, so the diffusion-weighted images no longer register to one another.
Ghosting and geometric warping in echo-planar readouts.
b-value and ADC error, because the actual gradient waveform (and thus the actual b-value) departs from the nominal one.

Scanners counter this with pre-emphasis: the gradient amplifier is driven with a shaped, overdriven waveform whose decay is tuned to cancel the eddy-current response so the net field matches the request. Pre-emphasis time constants are set at installation and can drift. Modern diffusion reconstructions add eddy-current correction (affine co-registration of the diffusion volumes) on top of the hardware compensation. Because the biases are both geometric and quantitative, eddy-current behavior belongs in the same QC conversation as calibration and linearity. ⁶⁷

A comparison of gradient failure modes and the tests that catch them

Gradient property	What fails	QC test that detects it	Typical tolerance / action level	Clinical signature
Gradient calibration (scaling)	Commanded ≠ actual amplitude near isocenter	ACR geometric accuracy (measured phantom lengths)	Within ≈ ±2 mm of true length	Systematically wrong measured sizes; ADC bias ≈ 2× the scaling error ¹⁶
Gradient linearity	Field departs from proportional-to-position toward FOV edges	Peripheral distortion / vendor gradient-nonlinearity check; large-FOV or grid phantom	Vendor-specified; error grows with distance from isocenter	Warped anatomy off-center; off-isocenter ADC bias over 5% if uncorrected ⁷⁸
Eddy-current compensation (pre-emphasis)	Residual transient field after gradient switching	Diffusion/EPI review for shear, shift, ghosting; vendor eddy tests	Qualitative: no residual shear/misregistration; b-value within spec	DWI misregistration, image shear, ADC error ⁶⁷
Slew rate / dB/dt	Amplifier cannot deliver commanded rise; or PNS limiting	Gradient output check; IEC operating-mode verification	IEC 60601-2-33 Normal / First-Level modes	Slower encoding, PNS at high dB/dt, acoustic noise ⁵
Slice position / thickness (gradient-dependent)	Slice-select gradient error	ACR slice position & thickness accuracy	Per ACR manual	Wrong slice location/thickness; partial-volume error ¹

Clinical Impact

Gradient errors are dangerous precisely because they are invisible on a casual read. A scanner with a 4% gradient scaling error still produces crisp, diagnostic-looking images — they are simply the wrong size and carry biased ADC. The consequences show up where numbers drive decisions:

Neuro-oncology and stroke. ADC thresholds distinguish tumor cellularity, grade restricted diffusion, and track treatment response. A systematic ADC bias from gradient miscalibration can shift a case across a decision boundary. ⁶⁷
MR simulation and stereotactic planning. Geometric fidelity is paramount when MRI defines a target or a stereotactic frame. Peripheral gradient nonlinearity that is fine for diagnosis can be unacceptable when a lesion sits off-isocenter, which is why distortion characterization is emphasized for MR-guided radiotherapy systems. ⁹
Multi-site and longitudinal quantitative studies. When ADC or volumetric measurements are compared across scanners or over time, differences in gradient calibration and transmission-chain hardware become confounders. A multi-site ACR-phantom program documented that a signal-transmission-chain upgrade measurably reduced geometric distortion along the slice direction and that vendor differences persisted in geometric measures. ⁹

The through-line: gradient QC is quality control for the numbers, not just the pictures.

Practical Optimization Tips

A defensible gradient QC program layers hardware checks, phantom tests, and quantitative validation.

1. Nail down geometric accuracy at isocenter first

Run the ACR geometric accuracy test on both the localizer and the ACR series, measuring the specified lengths and diameters. Investigate anything beyond about ±2 mm before trusting anything downstream. This is the fastest indicator of a gross calibration shift. ¹

2. Verify that distortion correction is actually on

Gradient-nonlinearity (spatial distortion) correction is a selectable option on most platforms and can be silently off for a given protocol. Confirm it is enabled for clinical protocols, especially large-FOV body, spine, and any MR-simulation or stereotactic work. Validate off-isocenter with a grid or large phantom, not only at center. ⁷⁸

3. Treat diffusion as a gradient stress test

Diffusion EPI pushes the gradients hardest and exposes both eddy-current and calibration problems. Review DWI/ADC for shear, shift, and ghosting, and periodically check ADC against a known diffusion phantom value (for example, the temperature-controlled ice-water standard). Because ADC error tracks about twice the gradient scaling error, ADC is a sensitive calibration monitor. ⁶

4. Check off-isocenter, not just center

Quantitative and geometric accuracy validated at isocenter degrades toward the periphery. For any protocol where the region of interest is routinely off-center, validate there. ⁷⁸

5. Re-verify after service

Gradient hardware replacement, coil service, cryogen events, and major software upgrades can all change calibration and pre-emphasis. Re-run geometric accuracy and a diffusion check after any of them, and after transmission-chain upgrades that vendors note can shift distortion. ⁹

Common pitfalls to avoid

Assuming the ACR phantom fully characterizes the gradients. It flags gross calibration near isocenter but does not fully probe peripheral nonlinearity or eddy-current behavior in demanding sequences.
Trusting isocenter results everywhere. Off-center accuracy must be shown, not assumed.
Ignoring ADC drift. A creeping ADC bias is often the first sign of a gradient calibration problem.
Leaving distortion correction off. A powerful correction does nothing if it is not enabled for the clinical protocol.
Skipping post-service QC. The most likely time for calibration to move is right after someone worked on the system.

Regulatory Considerations

MRI is non-ionizing, so it is not regulated as a radiation-producing machine — there is no MQSA-style federal mandate and, in the states DRPS serves, MRI falls outside the radiation-machine registration programs. That does not make MRI QC optional; it shifts the binding requirements to accreditation and hospital-standards bodies.

Key frameworks:

ACR MRI Accreditation Program and the ACR MRI Quality Control Manual (2015). These define the required QC tests — including geometric accuracy and the other gradient-dependent tests — the action levels, and the annual medical physicist evaluation. ¹²
AAPM Report No. 100, "Acceptance Testing and Quality Assurance Procedures for Magnetic Resonance Imaging Facilities." Provides the acceptance-test framework in which gradient calibration and related performance are established at installation. ³
NEMA MS 2-2008 (R2020), "Determination of Two-Dimensional Geometric Distortion in Diagnostic Magnetic Resonance Images." The standardized method for characterizing geometric distortion, which is the visible consequence of gradient calibration and nonlinearity. ⁴
IEC 60601-2-33. Governs MR equipment safety, including the gradient rate-of-change (dB/dt) operating modes that bound slew rate to control peripheral nerve stimulation. ⁵

For non-ionizing modalities in the states DRPS serves, the practical drivers are ACR accreditation and Joint Commission expectations rather than a state radiation-machine rule; the ACR accreditation physics requirements guide covers how those obligations fit together. Facilities should document gradient QC — acceptance values, periodic geometric accuracy, distortion-correction settings, and any post-service re-verification — as part of the medical physicist's report through medical physics consulting.

Frequently Asked Questions (FAQs)

What is MRI gradient quality control?

MRI gradient quality control is the set of tests that confirm the gradient subsystem encodes space correctly: that gradient amplitude is calibrated, that gradient fields stay linear across the imaging volume, and that eddy currents are adequately compensated. It protects geometric accuracy and quantitative values such as the apparent diffusion coefficient.

How is gradient calibration different from geometric distortion correction?

Gradient calibration sets the true scaling between the commanded gradient and the physical field, so distances map correctly near isocenter. Geometric distortion correction is a reconstruction step that unwarps residual gradient nonlinearity toward the edges of the field of view. Calibration is a hardware/scaling issue; distortion correction is a software correction of the field's spatial shape.

Why do eddy currents matter for diffusion MRI?

Diffusion sequences use large, rapidly switched gradients that induce eddy currents in nearby conductors. Residual eddy currents shift, shear, and scale the diffusion-weighted images relative to each other, causing misregistration and errors in the calculated apparent diffusion coefficient. Pre-emphasis and post-processing corrections reduce these effects.

Can the ACR MRI phantom detect gradient problems?

Partly. The ACR geometric accuracy test measures known lengths in the phantom and flags gross gradient miscalibration near isocenter, and slice-position and slice-thickness tests are also gradient-dependent. But the phantom is small relative to a full field of view and does not fully characterize peripheral gradient nonlinearity or eddy-current behavior in demanding sequences such as diffusion EPI.

Does gradient miscalibration affect ADC and other quantitative MRI?

Yes. Because the diffusion b-value depends on the square of the gradient amplitude, a gradient scaling error propagates roughly twofold into the ADC. A 5 percent gradient error can bias ADC by roughly 10 percent, which is clinically meaningful when ADC thresholds guide diagnosis or treatment response.

What tolerance does the ACR use for geometric accuracy?

In the ACR MRI Quality Control Manual, measured phantom lengths are expected to agree with the true dimensions within about 2 mm. Exceeding that action level triggers investigation of gradient calibration, gradient linearity, and distortion correction settings before the scanner is used for accredited imaging.

How often should gradient performance be evaluated?

Geometric accuracy and related gradient-dependent tests are part of the weekly or periodic technologist QC and the annual medical physicist evaluation under the ACR program. Gradient calibration is also verified at acceptance testing and after any gradient hardware, coil, or major software service event.

Key Takeaways

The gradient system encodes space, so gradient errors become geometry errors and quantitative errors. They do not announce themselves — images still look diagnostic.
Calibration, linearity, and eddy-current compensation are three distinct failure modes. Each needs its own test; the ACR phantom mostly covers calibration near isocenter.
ADC error tracks about twice the gradient scaling error. A 5% gradient error biases ADC by roughly 10%, so ADC doubles as a sensitive calibration monitor. ⁶
Off-isocenter accuracy must be demonstrated, not assumed. Peripheral gradient nonlinearity can push ADC bias past 5% when correction is inadequate. ⁷⁸
Re-verify after any service event. Gradient hardware, coil, cryogen, and software changes can move calibration and pre-emphasis.
Document everything against ACR and AAPM frameworks. MRI's requirements come from accreditation, not a radiation-machine rule.

Conclusion

Gradient QC is easy to under-appreciate because gradients rarely fail loudly. A scanner with a mis-scaled gradient, unenabled distortion correction, or drifting pre-emphasis will keep producing sharp, confident-looking images that are the wrong size or carry biased ADC values. The job of gradient quality control is to make those silent errors visible — through geometric accuracy at isocenter, distortion validation at the periphery, and diffusion/ADC checks that stress the system the way clinical protocols do.

For any facility relying on MRI for quantitative or geometrically precise work — neuro-oncology, stroke, MR simulation, or longitudinal quantitative studies — gradient QC is where the credibility of the numbers is established. A program that treats calibration, linearity, and eddy currents as three separate questions, validates off-isocenter, and re-verifies after service will produce images whose geometry and quantitative values can be defended.

How DRPS Can Help

Diagnostic Radiation Physics Services performs MRI acceptance testing and annual physicist evaluations that go beyond the ACR score sheet to characterize the gradient subsystem: geometric accuracy at and away from isocenter, verification that gradient-nonlinearity correction is enabled and effective, ADC validation against known phantom values, and post-service re-verification. This work is delivered through MRI physics testing, accreditation support, and medical physics consulting.

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

Good gradient QC is what lets a clinician trust that an ADC value means what it says — and that a target defined on MRI is where the scanner says it is.

Related Resources

References

American College of Radiology. ACR MRI Quality Control Manual. Reston, VA: American College of Radiology; 2015. acr.org
American College of Radiology. MRI Accreditation Program Requirements. acr.org
Jackson EF, et al. AAPM Report No. 100: Acceptance Testing and Quality Assurance Procedures for Magnetic Resonance Imaging Facilities. College Park, MD: American Association of Physicists in Medicine; 2010. aapm.org
National Electrical Manufacturers Association. NEMA Standards Publication MS 2-2008 (R2020): Determination of Two-Dimensional Geometric Distortion in Diagnostic Magnetic Resonance Images. Rosslyn, VA: NEMA. nema.org
International Electrotechnical Commission. IEC 60601-2-33: Medical electrical equipment — Particular requirements for the basic safety and essential performance of magnetic resonance equipment for medical diagnosis. Geneva: IEC. iec.ch
Teh I, Maguire ML, Schneider JE. Efficient gradient calibration based on diffusion MRI. Magn Reson Med. 2017;77(1):170-179. doi:10.1002/mrm.26105. PubMed
Neri JP, Koff MF, Koch KM, Tan ET. Validating the accuracy of multispectral metal artifact suppressed diffusion-weighted imaging. Med Phys. 2022;49(10):6538-6546. doi:10.1002/mp.15925. PubMed
Yamamoto T, Fukunaga M, Sugawara SK, Hamano YH, Sadato N. Quantitative evaluations of geometrical distortion corrections in cortical surface-based analysis of high-resolution functional MRI data at 7T. J Magn Reson Imaging. 2021;53(4):1220-1234. doi:10.1002/jmri.27420. PubMed
Palesi F, Nigri A, Gianeri R, et al. MRI data quality assessment for the RIN — Neuroimaging Network using the ACR phantoms. Phys Med. 2022;104:93-100. doi:10.1016/j.ejmp.2022.10.008. PubMed
Golestani AM, Gee JM. A statistical approach to automated analysis of the low-contrast object detectability test for the large ACR MRI phantom. J Appl Clin Med Phys. 2025;26(7):e70173. doi:10.1002/acm2.70173. PubMed
Wang J, Yung J, Kadbi M, Hwang K, Ding Y, Ibbott GS. Assessment of image quality and scatter and leakage radiation of an integrated MR-LINAC system. Med Phys. 2018;45(3):1204-1209. doi:10.1002/mp.12767. PubMed