[vtkusers] ISO-Value of bones

Stephen R. Aylward Stephen.Aylward at Kitware.com
Thu Jul 5 09:46:58 EDT 2007

Hi Sandra,

CT images are ACQUIRED in Hounsfield units.  These units (unlike, for 
example, MR) are on a absolute scale that has physical meaning, i.e., 
the radiodensity of the material.   The radiodensity of air is -1000 HU, 
water is 0 HU, and bone is about +1000 HU.   There will be variations 
between and within patients as bone density changes, but 1000 HU is a 
good start.

CT images are typically STORED in dicom files using unsigned shorts. 
So, the negative recorded values of HU units cannot be directly 
represented in the DICOM files.   Therefore, two additional fields are 
also stored in the dicom files when they contain CT data: RescaleSlope 
and RescaleIntercept.  For CT images, these additional values are 
typically set to 1 and -1000.  That is, an offset of -1000 must be 
applied to the recorded values in the dicom files to convert them to the 
acquired HU values.

MetaImages store dicom's RescaleSlope and RescaleIntercept values in the 
tags IntensityScale and IntensityOffset.  However, it is up to the DICOM 
to MetaImage converter to correctly fill-in and use those values - some 
converters don't do this :(.   Also, until recently (last month) VTK was 
using a VERY old versions of metaImage that did not support those tags. 
   ITK's (www.itk.org) version (the 'official' version of ITK) has been 
supporting them for years.   Last month we synchronized ITK and VTK's 
versions of MetaImage.

So, if (the tags IntensityScale and IntensityOffset are in the text 
header of your MetaImage file and the element type is unsigned shorts) 
or if (the element is signed shorts), you should probably use 1000 as a 
guess for bone density.   Otherwise, you should probably use 2000 for bone.

To account for inter- and intra-patient variation, you may want to also 
follow with a simple connectedness filter from ITK.

Also, if this is a contrast study (gastro has been injected prior to the 
scan to accentuate vasculature), then some vessels may be nearly as 
bright as the bone...and without a pre-contrast study, your job becomes 
a nice research topic.


samo wrote:
> Sorry, I have to correct myself, I am not looking for the iso-value, I am
> looking for the contour values I have to set in the marching cubes
> algorithm. 
> samo wrote:
>> Hi 
>> I would like to o know whether it is possible to calculate the iso-value
>> of an image for the marching cubes algorithm. I have different MetaImage
>> files (which represent different CT-data-files) which I use in a program.
>> I contruct the bone surface via marching cubes. There, I set the value
>> with cubes->SetValue(0,value). The optimal value changes for every file.
>> Is there a possibility to calculate the optimal value for representing the
>> bones out of the information I get from the MetaImage-file?
>> Thanks for your help!
>> samo

Stephen R. Aylward, Ph.D.
Chief Medical Scientist
Kitware, Inc. - Chapel Hill Office
Phone: (518)371-3971 x300

More information about the vtkusers mailing list