CT Number (Hounsfield Unit) Calibration and Accuracy QC

CT number accuracy is the quantitative foundation of computed tomography: every density-based diagnosis, PET/CT and SPECT/CT attenuation correction, and CT-to-electron-density conversion depends on Hounsfield units staying accurate and uniform. A defensible CT number QC program verifies the water value, material accuracy, spatial uniformity, and linearity under documented conditions, then compares the results against manufacturer specifications, ACR accreditation limits, and AAPM guidance.¹²³

A CT scanner that produces beautiful-looking images can still be quantitatively wrong. Because the Hounsfield scale is a calibrated, normalized quantity, even a few-HU systematic offset can shift a fat-versus-fluid call, change a contrast-enhancement measurement, or perturb a radiation therapy dose calculation. This guide explains how CT numbers are defined, why they drift, what tolerances apply, and how to build and document a CT number QC program that holds up under accreditation review and inspection.²³⁴

Introduction

The Hounsfield unit (HU) is a linear transformation of the measured linear attenuation coefficient, anchored so that water reads 0 HU and air reads -1000 HU. This normalization is what makes CT numbers comparable across patients, scanners, and time — but only if the scanner's air and water calibrations remain valid.¹²

CT number QC is therefore not a cosmetic check. It is the test that confirms the scanner is still measuring attenuation on the scale the entire clinical and quantitative workflow assumes. When a CT number drifts, the failure is silent: the images still look diagnostic, but the underlying numbers no longer mean what the radiologist, the PET/CT attenuation-correction algorithm, or the treatment planning system believes they mean.³⁶

This article walks through the definition of the Hounsfield scale, the radiation sources of CT number error, the tolerances used in ACR accreditation and AAPM TG-66 acceptance testing, a worked HU calculation, the clinical consequences of drift, and the practical steps that make a CT number QC program defensible. DRPS provides this work as part of its CT physics testing and accreditation support services across Florida, Maryland, Virginia, Washington DC, California, and Nevada.

Topic Explanation

What is a CT number?

A CT number is a voxel-by-voxel measure of how strongly tissue attenuates the X-ray beam, normalized to water and scaled by 1000. Soft tissues cluster near 0 HU, fat is negative, air is near -1000 HU, and dense bone or iodinated contrast is strongly positive. The fixed anchoring of water and air is what gives the scale clinical meaning.¹²

A practical CT number QC program asks a few core questions:

• Does water read close to 0 HU under defined conditions?
• Do known materials (air, polyethylene, acrylic, bone-equivalent) read within their expected ranges?
• Is the CT number uniform across the field of view, or does it cup or dish from center to periphery?
• Is the relationship between material attenuation and reported CT number linear?
• Are the results stable over time, and do they hold after service?

These questions map directly to the standard tests: water/material accuracy, uniformity, and linearity. For facilities pursuing or maintaining accreditation, they also map to specific pass/fail criteria — see ACR accreditation physics requirements for how these tests fit the broader program.

Why CT number accuracy is a quantitative problem, not a cosmetic one

Unlike subjective "does the image look good" checks, CT number accuracy is intrinsically quantitative. The same scanner can pass a visual review and fail a numerical one. Three downstream uses make this matter:

• Density-based diagnosis. Thresholds such as fat versus soft tissue, simple cyst versus complex fluid, and contrast enhancement are read directly in HU.
• Attenuation correction. PET/CT and SPECT/CT convert CT numbers into attenuation maps; CT number errors propagate into quantitative uptake values. See SPECT/CT quality control for the hybrid-imaging side of this.
• Radiation therapy planning. CT numbers are converted to relative electron density or mass density for dose calculation, so calibration errors can shift delivered dose.⁶

Key Technical Principles

The Hounsfield scale

The CT number of a voxel with linear attenuation coefficient $\mu$ is defined relative to the attenuation coefficients of water and air:¹²

Because ${\mu }_{\text{air}}\approx 0$ at diagnostic energies, this is very nearly the more familiar form:

By construction, water maps to 0 HU and air maps to -1000 HU. The scale is therefore only as good as the scanner's knowledge of ${\mu }_{\text{water}}$ and ${\mu }_{\text{air}}$, which is established during air and water calibration scans.

Worked example: converting attenuation to HU

Suppose a calibration material has a measured linear attenuation coefficient of $\mu =0.2090{\text{cm}}^{-1}$ at the effective beam energy, where ${\mu }_{\text{water}}=0.1900{\text{cm}}^{-1}$ and ${\mu }_{\text{air}}\approx 0$. Then:

So a material only about 10% more attenuating than water reads near +100 HU. This sensitivity is exactly why a small calibration error matters: a 1% shift in the effective ${\mu }_{\text{water}}$ moves the whole scale by roughly 10 HU, enough to cross several clinical decision thresholds. The same arithmetic shows why iodinated contrast (a few percent iodine by mass) can read in the hundreds of HU.

Comparison of common CT number QC tests

The numbers above are representative starting points for orientation, not a substitute for the current ACR CT Quality Control Manual edition, the manufacturer's specification, or the facility's documented action levels. Always verify the controlling limits before declaring pass/fail.²³

Linearity and electron-density calibration

For quantitative applications — especially radiation therapy — CT number must be a stable, near-linear function of material attenuation (often expressed as relative electron density). A calibration curve is built by scanning a phantom with multiple inserts of known composition and fitting reported HU against known electron or mass density. A multi-institution audit found measurable deviations between tissue-equivalent inserts and true human-tissue values across phantom vendors and planning systems, which is why the calibration curve — not just the water point — must be checked.⁶

where $f$ is the scanner- and protocol-specific calibration relationship loaded into the treatment planning system. Errors in $f$ propagate directly into dose calculation, so the curve should be re-verified after any change that can shift CT numbers.

Clinical Impact

A CT number offset is a quantitative error that hides behind normal-looking images. Its impact depends on where the numbers are used.

• Diagnosis at thresholds. Adrenal adenoma characterization, hepatic steatosis quantification, renal-cyst classification, and contrast-enhancement measurements are all read in HU. A systematic 10–20 HU offset can move a finding across a decision boundary.
• Quantitative and hybrid imaging. In PET/CT and SPECT/CT, CT numbers seed the attenuation map. Calibration drift biases standardized uptake values and quantitative reads.
• Radiation therapy. CT-to-electron-density conversion errors can shift calculated dose, particularly in heterogeneous regions such as lung and bone.⁶
• Longitudinal follow-up. When a patient is imaged on different scanners or after recalibration, uncorrected CT number differences can masquerade as real change.

Because reconstruction choices affect CT numbers and noise, QC results must be interpreted in the context of the clinical protocol. Our guide to Siemens reconstruction kernels explains how kernel selection changes the appearance and statistics of the image, and CT protocol optimization covers how technique choices interact with image quality and dose.

Practical Optimization Tips

A defensible CT number QC program follows a consistent workflow.

1. Standardize the conditions

CT numbers depend on kV, kernel, field of view, phantom size, and reconstruction method. Define and document the QC conditions so results are comparable over time. Daily water checks should use the manufacturer phantom and the manufacturer-specified technique.

2. Center the phantom

Off-center positioning and partial truncation are among the most common causes of false CT number and uniformity failures. Use the scanner's centering aids and confirm the phantom is fully within the scan field of view.

3. Use consistent ROIs

Place region-of-interest measurements consistently — typically a center ROI plus peripheral ROIs at fixed clock positions for uniformity. Keep ROI size and location stable between sessions so trends are meaningful.

4. Trend, don't just pass/fail

Record the actual HU values, not just pass/fail. A water value drifting from +1 HU toward +4 HU over weeks is an early warning, even if it never crosses the action level. Trending is what turns QC from a checkbox into a predictive tool.

5. Re-verify after change

Repeat CT number accuracy, uniformity, and (for RT) the electron-density curve after tube replacement, detector service, calibration scans, or software/kernel changes. Independent QC programs have shown CT-number behavior can change after service and detector recalibration, so post-service verification is essential.⁵

Common pitfalls to avoid

• Checking only water. Water at 0 HU does not guarantee correct values at bone or contrast densities; check multiple materials and the linearity curve.⁶
• Ignoring uniformity. Cupping or dishing can bias peripheral measurements even when the central water value is perfect.
• Mixing protocols. Comparing today's kernel/kV to last quarter's different settings produces meaningless trends.
• Skipping post-service checks. Detector recalibration and tube changes can shift CT numbers; verify afterward.⁵
• Confusing noise with accuracy. A low-noise image can still have a biased mean HU; they are different metrics.

Regulatory Considerations

CT number accuracy QC sits inside the broader framework of CT performance testing, accreditation, and state radiation-machine regulation. The controlling documents include manufacturer specifications, accreditation requirements, and recognized physics guidance.

• AAPM Report No. 39 (TG-1) and AAPM Task Group 66 — establish CT acceptance testing and quality assurance expectations, including CT number accuracy, uniformity, noise, and linearity as core tests for acceptance and routine QA.³⁴
• ACR CT Quality Control Manual and ACR CT Accreditation — define the phantom-based tests, action levels, and physicist/technologist responsibilities used by accredited facilities. CT number accuracy and uniformity are explicit accreditation measurements.²
• ACR–AAPM Technical Standard for Diagnostic Medical Physics Performance Monitoring of CT — defines the qualified medical physicist's role in annual performance evaluation, including CT number and image-quality testing.⁷
• IEC 61223-3-5 — international constancy- and acceptance-testing methodology for CT, including CT number measurements.⁸
• FDA 21 CFR 1020.33 — federal performance standard for CT equipment (a radiation-producing machine regulated by FDA and the states), with reporting and information requirements.⁹

Jurisdiction note: CT scanners are X-ray-producing machines regulated by the FDA under 21 CFR 1020.33 and by state radiation-control programs — in DRPS's footprint, Florida, Maryland, Virginia, California, and Nevada administer their own radiation-machine rules, while Washington, DC follows its local program; none of this falls under NRC byproduct-material rules. A qualified or board-certified medical physicist typically performs the annual evaluation and documents CT number accuracy. For the broader compliance picture, see common radiation safety violations and how to avoid them and ACR accreditation physics requirements.

Frequently Asked Questions (FAQs)

What is a CT number?

A CT number, expressed in Hounsfield units (HU), is a normalized measure of a voxel's linear attenuation coefficient relative to water. By definition water is 0 HU and air is -1000 HU. CT numbers let radiologists and quantitative tools compare tissue density on a fixed, scanner-independent scale.

Why does CT number accuracy matter clinically?

HU values drive density-based diagnosis (for example, fat, fluid, calcification, and contrast-enhancement thresholds), CT-based attenuation correction in PET/CT and SPECT/CT, and relative electron density used in radiation therapy dose calculation. Drift in CT number can shift measurements and degrade quantitative accuracy.

What is the tolerance for the water CT number?

Daily and periodic QC commonly require the mean CT number of water to be near 0 HU, with many programs using an action level on the order of about ±5 HU. The exact limit should follow the manufacturer specification, the ACR CT Quality Control Manual, and the facility's documented program.

How often should CT number accuracy be checked?

Water CT number and basic uniformity are typically checked daily by technologists using the manufacturer phantom. A qualified medical physicist verifies CT number accuracy across multiple materials, uniformity, and linearity at acceptance testing and at least annually, and after major service or tube replacement.

What causes CT numbers to drift?

Common causes include detector gain drift, tube aging, miscalibration of air or water reference scans, beam-hardening correction errors, off-center positioning, partial phantom truncation, temperature changes, and reconstruction or kernel changes. Calibration scans and service correct most systematic drift.

Does CT number accuracy depend on the protocol?

Yes. CT numbers can vary with kV, reconstruction kernel, field of view, phantom size, and iterative or deep-learning reconstruction settings. QC should be performed under defined, documented conditions, and thresholds should account for the protocol actually used clinically.

Who should perform CT number calibration QC?

Routine checks are performed by technologists, but the comprehensive CT number accuracy, uniformity, and linearity evaluation is performed or reviewed by a qualified or board-certified medical physicist as part of acceptance testing, annual performance evaluation, and accreditation support.

Key Takeaways

• CT number accuracy is quantitative, not cosmetic. Images can look normal while the underlying HU values are biased.
• The scale is anchored to water and air. A small error in the effective ${\mu }_{\text{water}}$ shifts the entire HU scale by roughly 10 HU per 1% — enough to cross clinical thresholds.
• Test more than water. Material accuracy, uniformity, and linearity must all be checked; a correct water value does not guarantee correct bone or contrast values.
• Trend the numbers. Recording actual HU values reveals drift before it crosses an action level.
• Re-verify after service. Tube changes, detector recalibration, and kernel/software updates can shift CT numbers.
• Document against the controlling standards. Manufacturer specs, the ACR CT QC Manual, AAPM TG-66, and the ACR–AAPM technical standard define the criteria and responsibilities.

Conclusion

CT number calibration QC is the test that keeps computed tomography quantitative. Because the Hounsfield scale is a calibrated, normalized quantity, even modest drift can change a diagnosis at a threshold, bias an attenuation-corrected uptake value, or perturb a radiation therapy dose calculation — all while the images continue to look diagnostic. A defensible program checks water and material accuracy, uniformity, and linearity under documented conditions; trends the results; re-verifies after service; and aligns the action levels with the manufacturer specification, ACR accreditation, and AAPM guidance. Facilities that treat CT number QC as a quantitative discipline rather than a daily checkbox protect both diagnostic accuracy and downstream quantitative workflows.

How DRPS Can Help

Diagnostic Radiation Physics Services (DRPS) supports CT and hybrid-imaging facilities with CT physics testing, acceptance testing, annual performance evaluations, CT number and image-quality QC program design, electron-density calibration verification for radiation therapy, and accreditation support prepared by board-certified medical physicists. We help translate manufacturer specifications and ACR/AAPM requirements into a documented, trendable QC program that holds up under accreditation review and inspection.

DRPS serves facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, and Nevada. For program design and ongoing oversight, see our medical physicist consulting services.

Related Resources

• CT protocol optimization
• CTDIvol and DLP dose metrics explained
• Siemens reconstruction kernels
• Metal artifact reduction in CT
• ACR accreditation physics requirements
• SPECT/CT quality control
• CT physics testing
• Accreditation support

References

1. Hounsfield GN. Computerized transverse axial scanning (tomography): Part 1. Description of system. British Journal of Radiology. 1973;46(552):1016-1022. doi:10.1259/0007-1285-46-552-1016. doi.org
2. American College of Radiology. CT Quality Control Manual. Reston, VA: ACR. acr.org
3. American Association of Physicists in Medicine. Specification and Acceptance Testing of Computed Tomography Scanners. AAPM Report No. 39 (Task Group 1). New York: AIP; 1993. aapm.org
4. American Association of Physicists in Medicine. Quality Assurance for Computed-Tomography Simulators and the Computed-Tomography-Simulation Process: Report of AAPM Task Group No. 66. Medical Physics. 2003;30(10):2762-2792. doi:10.1118/1.1609271. doi.org
5. Nakahara S, Tachibana M, Watanabe Y. One-year analysis of Elekta CBCT image quality using NPS and MTF. Journal of Applied Clinical Medical Physics. 2016;17(3):211-222. doi:10.1120/jacmp.v17i3.6047. doi.org
6. Nakao M, Ozawa S, Miura H, et al. CT number calibration audit in photon radiation therapy. Medical Physics. 2024;51(3):1571-1582. doi:10.1002/mp.16887. doi.org
7. American College of Radiology. ACR–AAPM Technical Standard for Diagnostic Medical Physics Performance Monitoring of Computed Tomography (CT) Equipment. acr.org
8. International Electrotechnical Commission. IEC 61223-3-5: Evaluation and routine testing in medical imaging departments — Part 3-5: Acceptance and constancy tests — Imaging performance of computed tomography X-ray equipment. Geneva: IEC. iec.ch
9. U.S. Food and Drug Administration. 21 CFR 1020.33: Computed tomography (CT) equipment. ecfr.gov
10. National Institute of Standards and Technology. XCOM: Photon Cross Sections Database (linear attenuation coefficients). nist.gov