# Hounsfield units and CBCT: can you read bone density?

> CBCT grey values are not calibrated Hounsfield units. What causes the gap, what phantom and calibration studies found, and what a viewer's number means.

- URL: https://jawviewer.com/blog/hounsfield-units-cbct-bone-density
- Category: Measurements explained
- Published: 2026-11-21
- Updated: 2026-11-21
- Author: Jaw Viewer
- Reading time: 5 min
- Tags: cbct, measurements, image-quality

## Key points

- In a phantom study on 13 CBCT devices and 1 medical CT, correlation between CBCT grey values and CT numbers ranged from 0.7014 to 0.9996 across the full density range.
- After recalibration, the average error of CBCT voxel values in the medium-density range was between 35 and 1562, so a good correlation did not mean a usable absolute value.
- A 2015 review names limited field size, scattered radiation and reconstruction algorithms as reasons CBCT grey values vary, and says quantitative use should generally be avoided.
- A 2022 systematic review of 3 studies found that CBCT grey values cannot be converted to Hounsfield units without equipment calibration, a prediction model and a standard formula.

Not reliably. A medical CT stores Hounsfield units, a defined scale, while a dental CBCT stores grey values that vary with the machine, the field size and the reconstruction. Phantom studies find a good correlation between the two and still large errors when the CBCT number is used as an absolute value.[1]

The gap between the scales has prompted phantom tests and attempts at calibration. For an overview of image limits, see [What a CBCT scan shows](https://jawviewer.com/blog/what-a-cbct-scan-shows.md).

## What the number is on each kind of scan

On a medical CT, each voxel holds a Hounsfield unit value. On a CBCT, each voxel holds a grey value. The two look alike in a viewer, because both are shown as shades from black to white. The review on bone quality notes that density used to be estimated from Hounsfield units on multidetector CT. Differences between CT and CBCT make quantitative use of CBCT grey values difficult.[2]

Jaw Viewer shows the grey value under the cursor when you point at a spot (see [view and measure](https://jawviewer.com/guide/view-and-measure.md)). On a medical CT file the number is the Hounsfield unit stored in the file. On a CBCT file it is that scanner's grey value. The viewer does not convert it, so it is not a calibrated density. For how the two kinds of scanner differ in other ways, see [CBCT and medical CT compared](https://jawviewer.com/blog/cbct-vs-medical-ct-differences.md).

## Why do the numbers vary?

Review authors list three causes that come with the technique itself: the limited field size, a relatively high amount of scattered radiation, and the limits of current reconstruction algorithms.[2] In other words (our paraphrase), the beam covers a small block of the head, stray X-rays add to the signal, and the software that builds the slices does not hold one fixed scale.

The practical result is that one material can read differently in two scans. The same review notes that attempts have been made to correct for the variability, and concludes that the quantitative use of grey values in CBCT should generally be avoided at this time.[2] That is a statement from 2015 about the state of the technique, not a rule for any single scan.

## What did the phantom study test?

The largest comparison used a phantom, an acrylic (PMMA) block with inserts of known density: air, PMMA, three concentrations of hydroxyapatite (50, 100 and 200 mg per cm3) and aluminium. It was scanned on 13 CBCT devices and 1 medical CT device.[1]

13 CBCT devices

plus 1 medical CT, same phantom[1]

0.7014 to 0.9996

correlation with CT numbers, full density range[1]

35 to 1562

average error of voxel values, medium range, after recalibration[1]

In the medium-density range the correlation was lower, from 0.5620 to 0.9991.[1] The abstract gives the average error without a unit, as values on each device's own scale. The authors conclude that most devices correlated well with CT numbers overall, but that large errors appear when the grey values are used in a quantitative way.[1] They add that pseudo-Hounsfield units might be obtainable from certain CBCT devices.[1]

A high correlation tells you the numbers rise and fall together. It does not tell you they agree, which is why a high correlation and a large absolute error can sit in the same study.

## What do the reviews say?

Two systematic reviews ask the question from different sides. One asked whether CBCT grey values can be converted to Hounsfield units. Only 3 studies qualified. It concluded that conversion is not supported, because clinical studies with diagnostic capacity are missing. It also lists three steps needed to get pseudo-Hounsfield values: equipment calibration, a prediction model and a standard conversion formula, rather than reading the number straight from the software.[3]

The other asked whether the two scales correlate. It included 14 studies, 2 of them in the meta-analysis, and found a moderate positive correlation of r = 0.669 (95% CI 0.388 to 0.836).[4]

| Question | Studies | Finding |
| --- | --- | --- |
| Can grey values be converted to Hounsfield units? | 3 | Not supported by clinical studies[3] |
| Do grey values correlate with Hounsfield units? | 14 (2 pooled) | Moderate positive correlation, r = 0.669[4] |
| How far off are grey values on 13 devices? | 1 phantom study | Average error 35 to 1562 in the medium range[1] |

The reviews are consistent: numbers can move together without being interchangeable.

## Has anyone calibrated a CBCT?

One clinical study tried. Patients wore an acrylic bite plate holding a reference object made of aluminium, cortical-bone-equivalent material, trabecular-bone-equivalent material, PMMA and water-equivalent material during a routine CBCT scan. There were 31 scans on one device and 30 on another, 61 in all.[5] The researchers related the grey levels of the reference materials to their known attenuation coefficients and scaled the result to Hounsfield units.

The relation between grey level and attenuation was in general satisfactory, which made it possible to calculate Hounsfield units from the grey values. The weak point was the effective energy of the beam. Uncertainty in determining it gave unrealistic energies and significant variability in the calculated CT numbers, and a regression straight from grey levels to Hounsfield units at set energies was more consistent.[5] So a calibration object in the mouth worked as a method, but it needs a reference in every scan and a device-specific fit.

## Where does this leave bone quality?

The 2015 review adds that research has moved the idea of bone quality from density towards bone structure. As CBCT image quality improves, it can show trabecular patterns, which suggests that structural analysis methods from micro-CT and histology might be applied.[2] That is a research direction, not something a viewer's grey value measures.

> What this means when you see a number
>
> A grey value on a CBCT scan is good for comparing two places inside the same scan, for example to see that one area is brighter than another. Do not read it as a bone density figure, and do not compare it with a number from a different scanner or a different scan.

## How does this relate to what you see?

Brightness on screen also depends on the display window you choose, which changes how grey values map to black and white without changing the stored numbers. See [why a scan looks too bright or too dark](https://jawviewer.com/blog/window-level-scan-too-bright.md). Dense metal adds streaks that shift nearby values, covered in [streaks and blur on a CBCT](https://jawviewer.com/blog/cbct-artefacts-streaks-near-metal-motion.md). Distances behave differently from densities, and [how accurate CBCT measurements are](https://jawviewer.com/blog/how-accurate-are-cbct-measurements.md) covers those.

This is general information, not medical advice. Jaw Viewer is not a medical device. Talk to your surgeon, orthodontist or radiologist about your own case.

## Questions and answers

### Can CBCT measure bone density in Hounsfield units?

Not reliably. A 2022 systematic review found CBCT grey values cannot be converted to Hounsfield units without device calibration, a prediction model and a standard formula, and that clinical studies supporting conversion are missing.

### Why do CBCT grey values differ between scanners?

A 2015 review names the limited field size, a relatively high amount of scattered radiation and the reconstruction algorithms. A phantom study on 13 devices also found large errors when grey values were used quantitatively.

### Do CBCT grey values correlate with CT Hounsfield units?

Yes, positively. A meta-analysis of 2 studies found r = 0.669 (95% CI 0.388 to 0.836). A correlation means the numbers rise together, not that one equals the other.

### Does Jaw Viewer convert CBCT grey values to Hounsfield units?

No. It shows the grey value stored in the file under the cursor. On a medical CT that value is in Hounsfield units. On a CBCT it is the scanner's own grey value, not a calibrated density.

## Terms used

- **CT**: Computed tomography. X-ray images taken from many angles are combined into slices of the body, with values in Hounsfield units.
- **Voxel**: A three-dimensional pixel. A scan is a grid of voxels, each with one grey value.
- **Hounsfield units**: The CT scale for how strongly a voxel blocks X-rays, with water at 0 and air near -1000.
- **CBCT**: Cone beam computed tomography. A scanner that rotates once around the head and records a cone-shaped X-ray beam, common in dentistry and jaw surgery.
- **Grey value**: The brightness number stored for a voxel. On a CBCT scan it is not a calibrated measure of density.

## Sources

1. Pauwels R, Nackaerts O, Bellaiche N et al. (2013). Variability of dental cone beam CT grey values for density estimations. Br J Radiol, vol. 86, pp. 20120135. https://doi.org/10.1259/bjr.20120135 One phantom scanned on 13 CBCT devices and 1 multislice CT device.
2. Pauwels R, Jacobs R, Singer SR et al. (2015). CBCT-based bone quality assessment: are Hounsfield units applicable?. Dentomaxillofac Radiol, vol. 44, pp. 20140238. https://doi.org/10.1259/dmfr.20140238
3. Eguren M, Holguin A, Diaz K et al. (2022). Can gray values be converted to Hounsfield units? A systematic review. Dentomaxillofac Radiol, vol. 51, pp. 20210140. https://doi.org/10.1259/dmfr.20210140 3 studies included.
4. Selvaraj A, Jain RK, Nagi R et al. (2022). Correlation between gray values of cone-beam computed tomograms and Hounsfield units of computed tomograms: a systematic review and meta-analysis. Imaging Sci Dent, vol. 52, pp. 133-40. https://doi.org/10.5624/isd.20210274 14 studies in the review, 2 in the meta-analysis.
5. Reeves TE, Mah P, McDavid WD (2012). Deriving Hounsfield units using grey levels in cone beam CT: a clinical application. Dentomaxillofac Radiol, vol. 41, pp. 500-8. https://doi.org/10.1259/dmfr/31640433 61 clinical scans on two devices with an intraoral reference object.

---

Jaw Viewer (https://jawviewer.com). Page index for AI tools: https://jawviewer.com/llms.txt. Not a medical device.
