Showing posts with label Hounsfield. Show all posts
Showing posts with label Hounsfield. Show all posts

Thursday 13 March 2014

Regarding calibration in the dicom image using ImageJ

1. I assume you have these in DICOM format. In ImageJ it is simple to import a single DICOM image and the calibration is done for you (the DICOM header contains fields for slope and intercept (usually 1 and -1024)). Since you have a stack this is probably not much good. For importing tomographic studies I use the "Import Dicom sequence" plugin available here: http://www.iftm.de/telemedizin/dcmimex.htm

From what I remember it is not the most straight-forward plugin to install but will nicely import a sequence of DICOM images as a stack. It does not, however, seem to calibrate the gray levels into hounsfield units. To do this choose Analyse->Calibrate. Choose "Straight line" as the function, type -1024 in the left box and 0 in the right box. When you press OK you get a straight-line graph of the calibration and a label with straight line formula y = a + bx. a should be -1024 and b should be 1. If they are then we have a calibration to HU.

To demonstrate the HU calibration, move the cursor around the image and observe the "value" in the IJ status bar. The value is in HU and the gray level appears in brackets.

2. Images typically only contain 256 gray levels when displayed, even though the image may contain values of any number (eg CT from -1024 to ~32k). So gray levels have to be "binned" in an image, just like in a histogram. So the column labelled "level" is the gray level displayed in the image and the "bins" are demonstrated in the second column. The size of the bin is dictated by the min and max pixel levels.

I hope I pitched that at the right level. Enjoy ImageJ ;-)

Converting CT Data to Hounsfield Units

According to Wikipedia, the Hounsfield scale was invented in 1972 by Godfrey Newbold Hounsfield. His scale is a quantitative measure of radiodensity and is used to evaluate CAT scans. Pixels in an image obtained by CT scanning are displayed in terms of relative radiodensity. 

The pixel value is displayed according to the mean attenuation of the tissue that it corresponds to on a scale from -1024 to +3071 on the Hounsfield scale. Water has an attenuation of 0 Hounsfield units (HU) while air is -1000 HU, bone is typically +400 HU or greater and metallic implants are usually +1000 HU. 

To convert from the normal units found in CT data (a typical data set ranges from 0 to 4000 or so) you have to apply a linear transformation of the data. The equation is:
   hu = pixel_value * slope + intercept
 
The real question is where do you find the slope and intercept used in the conversion?
Normally, these values are stored in the DICOM file itself. The tags are generally called the Rescale Slope and Rescale Intercept, and typically have values of 1 and -1024, respectively.

To show you how to obtain these values, I downloaded a sample CT data set, named CT-MONO2-16-ankle.dcm. This file was created on a GE Medical Systems scanner. After unpacking the compressed file, and adding a dcm file extension to the name (a convenience), I opened the file and dumped the elements to the display.
   IDL> dicomObj = Obj_New('IDLffDICOM', 'CT-MONO2-16-ankle.dcm')
   IDL> dicomObj -> DumpElements
      0 : (0002,0000) : UL : META Group Length : 4 : 188 
      1 : (0002,0001) : OB : META File Meta Information Version : 2 : 0 1 
      2 : (0002,0002) : UI : META Media Stored SOP Class UID : 26 : 1.2.840.10008.5.1.4.1.1.7
     ...
     ...
     50 : (0028,1052) : DS : IMG Rescale Intercept : 6 : -1024 
     51 : (0028,1053) : DS : IMG Rescale Slope : 2 : 1 
     52 : (0028,1054) : LO : IMG Rescale Type : 2 : US
     53 : (7FE0,0000) : UL : PXL Group Length : 4 : 524296 
     54 : (7FE0,0010) : OW : PXL Pixel Data : 524288 : 4080 4080 4080 4080 4080 ...
I found the Rescale Slope and Rescale Intercept as elements 51 and 50. As expected, they had values of 1 and -1024.
Next, I read the data from the DICOM file, and applied the transformation.
 
   IDL> imagePtr = (dicomObj -> GetValue('7FE0'x, '0010'x))[0]
   IDL> MinMax, *imagePtr 
        32   4080
   IDL> image_hu = *imagePtr * 1 + (-1024)
   IDL> MinMax, image_hu
        -992   3056
This image will appear upside down on my display, so I want to reverse the Y direction.
   IDL> image_hu = Reverse(image_hu, 2)
If I just want to see the bone structure (probably a good idea with this ankle image), I can display it like this.
   TV, BytScl(image_hu, Min=600, Max=3000)
The CT image displayed in Hounsfield units.
The bone structure of the CT angle image, displayed in Hounsfield units.
 
Be sure to clean up your pointers and objects.
   IDL> Ptr_Free, imagePtr
   IDL> Obj_Destroy, dicomObj