MRI Diffusion ADC Quantitative QC

Introduction

A diffusion-weighted MRI number is only a biomarker if it is reproducible — and reproducibility has to be measured, not assumed. A quantitative DWI quality-control (QC) program uses a temperature-controlled diffusion phantom with known apparent diffusion coefficient (ADC) values, a standardized sequence with verified b-values, and defined bias and repeatability tolerances to prove that an ADC of, say, $1.0 \times 1 0^{- 3} mm^{2} / s$ means the same thing on Monday as it did last quarter, and the same thing on your scanner as on the one across town. ^{1, 2}

Diffusion-weighted imaging (DWI) has moved well beyond stroke detection. Radiologists and oncologists now read the ADC as a quantitative marker of tumor cellularity and treatment response, and clinical trials increasingly use ADC change as an endpoint. ^{1, 3} But the moment an ADC value is reported as a number, the number carries an implicit promise: that it is accurate and stable. Standard MRI accreditation QC — geometric accuracy, signal-to-noise ratio (SNR), uniformity — does not test that promise. ⁴

This guide explains how to build a quantitative DWI/ADC QC program: what the ADC actually measures, why temperature and b-value accuracy dominate the error budget, what phantom and tolerances to use, and how the program fits alongside ACR MRI accreditation and the annual medical-physicist evaluation. DRPS provides this work as part of its MRI physics testing and accreditation support services across Florida, Maryland, Virginia, Washington DC, California, and Nevada.

Topic Explanation

What does DWI actually measure?

Diffusion-weighted imaging measures the microscopic, thermally driven motion of water molecules within tissue, and expresses it as the apparent diffusion coefficient. In free water, molecules diffuse randomly; in tissue, cell membranes, macromolecules, and organelles restrict that motion. Densely cellular tissue — many tumors — restricts diffusion and produces a low ADC, while cystic or necrotic tissue diffuses more freely and produces a high ADC. ^{1, 3}

The word "apparent" matters. The scanner does not measure a pure physical diffusion constant; it measures an effective value that folds in restriction, perfusion at low b-values, and any imperfection in the acquisition. That is exactly why QC is required: several of those imperfections are under the facility's control, and the rest need to be characterized so they do not masquerade as biology.

For related quantitative-imaging QC in other modalities, see our companion work on task-based CT image quality and the noise power spectrum, which shares the same philosophy: a reported number needs a measured uncertainty.

What is the ADC, numerically?

DWI acquires at least two images with different diffusion weightings, described by the b-value (units of s/mm²). Signal intensity falls off with b-value according to a mono-exponential model. With images at two b-values, the ADC is the slope of the log-signal versus b-value:

S (b) = S_{0} e^{- b \cdot ADC}

ADC = \frac{ln ( S ( b _{1} ) ) - ln ( S ( b _{2} ) )}{b _{2} - b _{1}} = \frac{1}{b _{2} - b _{1}} ln (\frac{S ( b _{1} )}{S ( b _{2} )})

Because ADC is computed from a ratio of signals across a difference of b-values, two error sources dominate: the accuracy of the applied b-values (the denominator) and the fidelity of the signal (the numerator, which is sensitive to eddy currents, gradient nonlinearity, and noise). A QC program is, at heart, a way to bound both. ^{2, 5}

The phantom problem, and how NIST/QIBA solved it

A quantitative phantom needs materials with known, stable diffusion values across the clinically relevant range. Water alone is not enough, and — critically — diffusion is strongly temperature dependent. The National Institute of Standards and Technology (NIST), with the National Cancer Institute and the RSNA Quantitative Imaging Biomarkers Alliance (QIBA), developed a diffusion phantom that solves both problems: thirteen vials of pure water and polyvinylpyrrolidone (PVP) solutions at 0, 10, 20, 30, 40, and 50% by mass, all held in an ice-water bath near 0 °C so the reference ADC values are defined at a single reproducible temperature. Higher PVP concentration yields lower ADC, spanning roughly $0.1$ to $1.1 \times 1 0^{- 3} mm^{2} / s$ . ^{2, 6}

Key Technical Principles

The ADC error budget

The following table summarizes the main contributors to ADC error and how each is controlled in a QC program. This is the core of why "the images look fine" is not the same as "the ADC is correct."

Error source	Effect on ADC	Primary control	QC check
Temperature of phantom	~2–3% per °C; dominates if uncontrolled	Ice-water bath, equilibration, measured temperature	Confirm 0 °C equilibration before scanning ^{2, 6}
b-value inaccuracy	Directly scales ADC (denominator)	Vendor calibration; verify against known-ADC vials	Bias vs. reference ADC per vial ^{2, 5}
Gradient nonlinearity	Spatially varying bias, worse off isocenter	Keep phantom/ROI near isocenter; apply correction	Position dependence of ADC ⁵
Eddy currents / distortion	Misregistration between b-value images	Bipolar/twice-refocused diffusion prep	Visual + geometric check ²
Low SNR at high b-value	Noise-floor bias, ADC underestimate	Adequate averaging; avoid excessive b	Repeatability (wCV) at low-ADC vials ^{1, 2}
ROI placement / drift	Adds variance run to run	Fixed ROI template, central slice	Repeatability across sessions ¹

Worked ADC example

Suppose a two-b-value acquisition ( $b_{1} = 0$ and $b_{2} = 1000 s/mm^{2}$ ) yields a mean signal of $S (0) = 1000$ and $S (1000) = 300$ in a phantom vial. The ADC is:

ADC = \frac{1}{1000 - 0} ln (\frac{1000}{300}) = \frac{ln ( 3.33 )}{1000} = \frac{1.204}{1000} = 1.20 \times 1 0^{- 3} mm^{2} / s

Now suppose the scanner's actual diffusion weighting was 3% low — the true b-value was 970, not 1000 — but the reconstruction assumed 1000. The reported ADC would be computed with the wrong denominator, biasing the result by roughly the same 3%. Nothing in the image would look wrong; only a known-value phantom exposes it. This is the reason b-value accuracy sits at the center of the error budget. ^{2, 5}

Bias and repeatability, defined

Two distinct quantities describe ADC performance, and a QC program must report both:

Bias — the systematic difference between the measured ADC and the known reference ADC, usually expressed as a percentage of the reference. Bias is an accuracy statement.
Repeatability / reproducibility — the random scatter of repeated measurements, usually expressed as a within-scanner repeatability coefficient of variation (wCV, same scanner over time) or a cross-scanner reproducibility CV (between scanners). These are precision statements.

The QIBA DWI Profile sets technical-performance claims for these quantities. In recent multi-site phantom studies using the standardized QIBA sequence, intra-scanner and inter-scanner CVs stayed within roughly 2.2%, and bias within about ±3.6% of the known vial value, with accuracy and precision both degrading for the lowest-diffusivity (lowest-ADC) vials. ^{2, 7} Those numbers are useful acceptance anchors, but each site should confirm the current Profile claim and apply it to its own scanner and sequence.

Clinical Impact

When ADC is reproducible, a change in ADC can be trusted to reflect a change in the patient; when it is not, an apparent change may be nothing more than scanner drift. That distinction determines whether ADC is safe to act on.

Tumor characterization. Low ADC suggests high cellularity. If a scanner's ADC bias drifts by 10%, a borderline lesion can be pushed across a decision threshold. ^{1, 3}
Treatment response. In oncology, a rising ADC often indicates successful therapy (loss of cellularity). In the multicenter ACRIN 6698 breast trial, the change in tumor ADC after neoadjuvant chemotherapy predicted pathologic complete response — but only because acquisition was standardized across ten sites. ⁹ A response endpoint of, for example, a 15% ADC increase is meaningless if the measurement's own reproducibility is 15%. The measurement uncertainty must be smaller than the effect you want to detect. ¹
Multi-scanner and longitudinal care. Patients are imaged on whichever scanner is available. Without cross-scanner reproducibility QC, a patient scanned on Scanner A at baseline and Scanner B at follow-up may show an "ADC change" that is pure inter-system bias. ^{1, 2}
Trial eligibility. Sites that join cooperative-group or industry oncology trials are frequently required to pass QIBA-style phantom qualification before enrolling patients. A standing DWI QC program turns that from a scramble into a formality. ¹

Practical Optimization Tips

Standardize the sequence first

Freeze a single, documented DWI QC sequence and do not let it drift with protocol edits. Match it to the QIBA Profile guidance where possible: a small number of well-separated b-values, adequate SNR, and diffusion encoding that minimizes eddy-current distortion. Record every parameter (b-values, TR/TE, resolution, averages, gradient mode) so a future measurement is truly comparable. ²

Respect the temperature

Fill and equilibrate the phantom in ice water, and confirm it has reached thermal equilibrium before scanning — a warm phantom will read high. Do not start immediately after moving the phantom from a warm hallway. Log the fact that equilibration was confirmed; temperature is the single largest uncontrolled error if you skip this step. ^{2, 6}

Keep the measurement near isocenter

Gradient nonlinearity biases ADC increasingly toward the edges of the field of view. Center the phantom, analyze central slices, and place ROIs consistently. If you must measure off-center, apply the vendor's gradient-nonlinearity correction and note it. ⁵

Analyze with a fixed template

Use the same ROI size and location every time, on the same vials, and compute both bias (versus the known values) and repeatability across sessions. Trend the results on a control chart. A gradual bias creep after a gradient service event is exactly the kind of signal a trended program catches and an ad-hoc check misses.

Re-baseline after changes

Repeat the full ADC QC after any gradient hardware service, coil change, or software/reconstruction upgrade. Vendors can change diffusion-preparation or gradient calibration silently in an update, and a re-baseline is how you keep the historical trend meaningful. ²

Common pitfalls to avoid

Trusting a pretty image. ADC bias is invisible on the image; only known-value phantoms reveal it.
Skipping temperature control. A few degrees of error can exceed your entire tolerance budget.
Using too many high b-values in QC. Excessive diffusion weighting drives signal into the noise floor and biases ADC low, especially in low-ADC vials.
Changing the sequence between sessions. If the QC sequence drifts, the trend is worthless.
Confusing accuracy with precision. A scanner can be precisely wrong (tight repeatability, large bias). Report both.

Regulatory Considerations

Quantitative DWI QC sits on top of, not inside, the mandatory MRI QC framework — so a facility must understand where accreditation ends and quantitative-imaging QC begins. Unlike mammography under MQSA, general diagnostic MRI has no federal quality mandate, and because MRI is non-ionizing it also falls outside state radiation-machine programs. The binding requirements come from accreditation and hospital standards. ⁴

ACR MRI Accreditation Program. Requires an annual MRI performance evaluation by a qualified medical physicist or MR scientist, plus a technologist QC program, covering geometric accuracy, spatial resolution, low-contrast detectability, SNR, artifact, and image uniformity. It does not, by itself, mandate a dedicated quantitative-ADC test — which is precisely the gap a DWI QC program fills for facilities that report ADC. ⁴
ACR–AAPM Technical Standard for Diagnostic MRI. Defines the medical physicist's role and the scope of the annual evaluation; a quantitative-imaging QC program is a natural extension of that scope where ADC is used clinically. ¹⁰
QIBA Diffusion-Weighted MRI Profile. The RSNA/QIBA Profile is a voluntary technical-performance standard, not a regulation, but it is the reference for bias and repeatability claims and for trial qualification. ²
IEC 62464-1 provides standardized definitions and test methods for MR image-quality parameters that underpin acceptance and constancy testing. ⁸

Facilities pursuing or maintaining accreditation should coordinate the DWI QC program with the annual physicist evaluation and with MRI physics testing, accreditation support, and broader medical physics consulting. For the baseline accreditation testing this program extends, see our guides on ACR MRI phantom QC and ACR accreditation physics requirements.

Frequently Asked Questions (FAQs)

What is a quantitative DWI QC program?

It is a program that verifies the apparent diffusion coefficient (ADC) produced by a scanner is accurate, repeatable across time, and reproducible across scanners. It uses a temperature-controlled diffusion phantom with known ADC values, fixed b-values, a standardized sequence, and defined bias and repeatability tolerances such as those in the QIBA DWI Profile.

Why does temperature matter so much for ADC?

Water diffusion is strongly temperature dependent, changing on the order of a few percent per degree Celsius. A diffusion phantom must be measured at a known, stable temperature, which is why the NIST/QIBA phantom is equilibrated in an ice-water bath near 0 degrees Celsius so its reference ADC values are defined.

What ADC value should pure water read at 0 degrees Celsius?

The reference ADC of pure water at 0 degrees Celsius is approximately 1.1 x 10^-3 mm^2/s. The NIST/QIBA diffusion phantom uses this value, along with polyvinylpyrrolidone (PVP) solutions that produce lower ADC values, to cover the clinically relevant range down to roughly 0.1 x 10^-3 mm^2/s.

How are b-values related to ADC?

b-value is the diffusion weighting, set by gradient strength, duration, and spacing. ADC is calculated from the change in signal between at least two b-values using a mono-exponential fit. If the applied b-values are inaccurate, or too few, the ADC will be biased even when the images look normal.

What tolerances define an acceptable ADC?

The QIBA DWI Profile sets technical-performance claims for bias and repeatability. In multi-site phantom work, cross-scanner reproducibility and within-scanner repeatability coefficients of variation on the order of 2 percent, and bias within a few percent of the known value, are typical acceptance targets for the standardized sequence at higher ADC values.

Is DWI QC required for ACR MRI accreditation?

ACR MRI accreditation and the ACR–AAPM technical standard require an annual performance evaluation by a qualified medical physicist and routine technologist QC, but they do not mandate a dedicated quantitative-ADC test for general accreditation. Facilities that report ADC as a quantitative result, or that join oncology trials, should add QIBA-style ADC QC on top of the accreditation baseline.

How often should ADC QC be performed?

A practical cadence is a baseline at acceptance, a repeat after any gradient or software change, and periodic checks (for example quarterly or semiannually) plus the annual physicist evaluation. Sites enrolled in quantitative trials often follow the trial's phantom schedule, which can be monthly.

Key Takeaways

ADC is a ratio across a b-value difference, so its two dominant error sources are b-value accuracy and signal fidelity — neither of which is visible on a normal-looking image.
Temperature is the largest uncontrolled error. The NIST/QIBA phantom is defined in an ice-water bath at 0 °C, where pure water reads about $1.1 \times 1 0^{- 3} mm^{2} / s$ .
Report bias and repeatability separately. A scanner can be precisely wrong; accuracy and precision are different claims.
QIBA Profile tolerances — CVs on the order of 2% and bias within a few percent for the standardized sequence at higher ADC — are practical acceptance anchors, worse at low ADC.
Quantitative DWI QC extends accreditation; it is not part of it. ACR and the ACR–AAPM standard require an annual physicist evaluation but not a dedicated ADC test.
Re-baseline after gradient, coil, or software changes, and trend results on a control chart so drift is caught early.

Conclusion

Reporting an ADC value is a quantitative claim, and quantitative claims need quantitative QC. A program built around a temperature-controlled diffusion phantom, a frozen standardized sequence, verified b-values, and QIBA-referenced bias and repeatability tolerances turns the ADC from a picture into a measurement with a known uncertainty. That is what lets a radiologist compare today's scan to last quarter's, and this scanner to the one down the hall, without wondering whether a change is the patient or the machine.

The medical physicist's job here is not to add bureaucracy but to protect the meaning of a number that increasingly drives clinical decisions. Facilities that report ADC — and especially those entering oncology trials — should treat quantitative DWI QC as a standing program, coordinated with their annual MRI evaluation, not a one-time exercise.

How DRPS Can Help

Diagnostic Radiation Physics Services helps MRI facilities translate quantitative-imaging requirements into practical, documented QC. For DWI programs, this may include diffusion-phantom acceptance and constancy testing, b-value and gradient-nonlinearity assessment, QIBA-referenced bias and repeatability analysis, control-chart trending, trial-qualification support, and integration with MRI physics testing, the annual physicist evaluation, and accreditation support.

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

A quantitative imaging program is only as trustworthy as the QC behind it — the goal is an ADC your radiologists can act on with confidence.

Related Resources

References

Shukla-Dave A, Obuchowski NA, Chenevert TL, et al. Quantitative Imaging Biomarkers Alliance (QIBA) recommendations for improved precision of DWI and DCE-MRI derived biomarkers in multicenter oncology trials. J Magn Reson Imaging. 2019;49(7):e101-e121. doi:10.1002/jmri.26518. PubMed
Radiological Society of North America, Quantitative Imaging Biomarkers Alliance. QIBA Profile: Diffusion-Weighted Magnetic Resonance Imaging (DWI). qibawiki.rsna.org
Padhani AR, Liu G, Koh DM, et al. Diffusion-weighted magnetic resonance imaging as a cancer biomarker: consensus and recommendations. Neoplasia. 2009;11(2):102-125. doi:10.1593/neo.81328. PubMed
American College of Radiology. MRI Accreditation Program Requirements. acr.org
Malyarenko D, Galbán CJ, Londy FJ, et al. Multi-system repeatability and reproducibility of apparent diffusion coefficient measurement using an ice-water phantom. J Magn Reson Imaging. 2013;37(5):1238-1246. doi:10.1002/jmri.23825. PubMed
National Institute of Standards and Technology. MRI diffusion phantom. nist.gov
Carr ME, Keenan KE, Beavan M, et al. Quantifying multi-institutional ADC measurement variability of 1.5 T MR-Linacs: A phantom and in vivo study. Med Phys. 2025;52(6):4120-4133. doi:10.1002/mp.17739. PubMed
International Electrotechnical Commission. IEC 62464-1:2018 Magnetic resonance equipment for medical imaging — Part 1: Determination of essential image quality parameters. iec.ch
Partridge SC, Zhang Z, Newitt DC, et al. Diffusion-weighted MRI findings predict pathologic response in neoadjuvant treatment of breast cancer: the ACRIN 6698 multicenter trial. Radiology. 2018;289(3):618-627. doi:10.1148/radiol.2018180273. PubMed
American College of Radiology, American Association of Physicists in Medicine. ACR–AAPM Technical Standard for Diagnostic Medical Physics Performance Monitoring of Magnetic Resonance Imaging (MRI) Equipment. acr.org