This is an important question but unfortunately with no definite answer in the literature.

But let' me share these points with you:.

1. There is broadly two types of bones: cancellous and compact bone, and the expected range of HU densities for each type is expected to be different.

2. There are several factors that can influence bone density including age, (and mild age-related osteopenia), and gender.

3. There are different ranges of CT densities of bone in the literature.

4. My experience in measuring CT densities of abnormal ossifications/calcifications of tendons gave me results as low as 300 HU; the upper limit of scale was about 2000 HU.

5. There are softwares that may help in segmentation of CT images and discrimination of the tissues based on their CT densities (HU). I use (InVesalius) which applies a mask overlying the tissues with the chosen HU range.

6. In (InVesalius), there are ranges of densities for each type of bone according to the age:

e.g. Compact bone in adult is 660 to 1998 HU. This range differs in children, and in cancellous bone.

7. You may try on control group and know the ranges of your studied cases.

Then you may use (custom range) of HU that is suitable most for your cases.

I hope this helps

Best

You will see that there will be a variety of HU values between patients, (1) because of the issues Sahar mentioned and (2) because of the polychromatic nature of the X-rays that are used in CT/CBCT.. Nonetheless, rather than changing the threshold manually all the time you could also use thresholding using various global (histogram-derived) and local (adaptive) methods (see more here: http://www.mecourse.com/landinig/software/autothreshold/autothreshold.html). ImageJ/Fiji would give you that option, don't know about software provided by the producer of the CT

Sahar N Saleem and Michael Sandholzer above give very good answers. I would further add that many software viewers for CT (and MRI) scans offer preset HUs for bone, brain tissue, muscular tissue and so on. Selecting any of these will instantly invoke the preset HUs and highlight the chosen material in the display. That is a valuable asset which saves much time, and in the majority of cases the HU’s embedded in the presets are good enough. However I have observed, and I’m sure many others have too, that the detailed appearance in CT scans of the same region in different patients can appear slightly different even with the same preset HUs applied. There are likely to be many reasons for this, some of which are mentioned by Saleem and Sandholzer above. For most users that difference is not an issue, but perhaps in your case it is. Either way, I think the inevitable answer to your question is a definite maybe :-)

To add to the other good answers provided already:

(1) If it is practical for your scanning techniques, you could include a calibration phantom with your CT scans. These phantoms have rods that contain known concentrations of calcium hydroxyapatite, allowing a post-processing conversion of HU to equivalent BMD. Applying a given threshold in BMD values rather than HU may lead to a more accurate and consistent distinction between bone and non-bone material, and between cancellous and cortical bone.

(2) Another consideration is the spatial resolution of your scans, which will impact the amount of spatial averaging of bone and non-bone material, particularly at the boundaries between bone and soft tissue or between bone and air. As spatial resolution decreases (i.e. voxel size gets larger), choosing any threshold will lead to an unavoidable trade-off between sensitivity and specificity.

Tom Hangartner and colleagues at Wright State University have written rather extensively about these trade-offs, e.g.

Hangartner TN (2007). Thresholding technique for accurate analysis of density and geometry in QCT, pQCT, and microCT images. J Musculoskelet Neuronal Ineract 7(1):9-16.

Good luck!

When you look at definition of HU values, you can see that the values are really defined only for water (by definition 0) and for vacuum (-1000). For other tissues/materials the numbers depend on the beam quality, because linear attenuation coefficient of radiation depends not only on the material, but also on the radiation. You will notice some differences between HU values for a single sample of bone tissue depending on choice of tube voltage (kVp) and filtration (some scanners use different filters for head and body) of the beam on your CT scanner.

Also, as Michael Sandholzer wrote above, there is a problem of the "polychromatic nature of the X-rays". The x-ray beam has a quite wide spectrum. When the beam passes through patient it is attenuated, and the spectrum changes. Lower energies are attenuated more than higher energies, so the spectrum gets "harder", an linear attenuation cefficient of bone also changes. So you can get different HU values for the same sample of bone tissue placed in air, in a thin patient, and in a thick patient, as the beam that reaches the bone is slightly different in each case.

The dependence of HU values on the beam quality will certainly add some uncertainty to segmentation.

In order to overcome the polychromatic issues of the X-ray beam, one may use dual energy CT. In this case each material (in our case the calcium) can be segmented from the other material and you can also get a quantitative measurement if all the preliminary work is done... One point to also mention is that the amount of scatter will differ depending the size of the patient and the amount of dense tissue.

Don't forget the instrumental issues. The scanner shoudl be periodically tested to check the correspondence of visualized HU and known HU. This is done with a phantom whose different plastic inserts have known HU.

The segmentation of bones can be surprisingly tricky.

If your purpose is visualization, e.g. 3D rendering, you might get away with a threshold and may be some pre- and postprocessing.

If you are interested in quantitative analysis, things get complicated very quickly. Nearly all long bones have areas of thin cortex in the vicinity of joints, where thresholding will fail. Vertebral bodies have only thin cortical bone, so that a threshold will generate huge errors especially in elderly people with decreased BMD, or with aortic/vascular calcification.

Using a BMD calibration phantom as David Gutekunst mentioned is essential for any quantitative analysis of BMD, and will improve the stability of your thresholds.

You can have a look at some of my papers, we used a vertebral template and active shape deformation for the precise segmentation of vertebra. The work of Klaus Engelke et al. is very good, also for other anatomy sites.

As some other people have already mentioned HU is depended to Kv of ur scanner. Using dual source CT helps u to measure HU more accurately. Partial volume measurement may also cause some errors if u choose your sample volume to large or include cancellous and cortical bone in same measurement

Hi, If you interested in the HU in CBCT and MSCT and potentially how do these relate I have with some other guys a good article on my profile. Internal validation .... cadaver study. Dr Sandholzer has also contributed to this article kindly,

I am thankful to all putting light on HU numbers for segmentation of tissues. I feel the Histogram will definitely help in understanding the attenuation pattern for individual and then we can selct the individual tissue for segmentation may it be bone or other tisuue then customize it by visual cues if all agree