That's odd. I've been using Image J for lung CT images and it gives Hounsfield units if you are using DICOM images. You can even use it for segmentation based on threshold values of HU.

ImageJ does calibrate the imported DICOM image based on information from the header file (slope and intercept). The latter fields are normally recorded as 1 (slope) and -1024 (intercept). ImageJ automatically displays the Hounsfield units  in the status bar when placing the curser on the image, for example:

x=404, y=218, value= -896.00 (31872)

where -896 is your HU at the pixel location p(x,y).

The assumption with the conversion is that CT intensity values are consistent. With a given machine, radio-density in a voxel of an examined sample has to be reproducible. Among the factors that hinder arriving at a consistent reading (if a machine is not calibrated) are: the geometry (beam not orthogonal to the axis of rotation), scattered radiation, beam hardening, metal artifacts, etc.

The Hounsfield units (HU) conversion is a linear transformation such as:

HU = Gray_Value * slope + intercept

(slope and intercept can be retrieved from their corresponding fields in the image header file as mentioned earlier).

Dear Marwan,

I would like to add few more points.

Pixel intensity comes in to picture only for the visualization purpose i.e the converted voxel intensity based on the desired window settings we use for 2D visualization.

The stored value in DICOM file are basically the attenuation coefficients i.e CT numbers (DICOM tag (7FE0, 0010) - pixel data). These are converted to a linear scale (Hounsfield unit scale) by using HU = CTNumber (7FE0,0010) * rescale slope (0028, 1053) + y intercept (0028, 1052). These HU values have to be converted to pixel intensity f(x,y) using f(x,y) = ((HU-P1)*pow(2,i-1))/W, where P1 is the window left border, i is the radiometric resolution and W is the window width. 

I would like to call this as pixel intensity (the final display quantity). In your case, with user interaction you can display both f(x,y) as well as HU(x,y) based on your pixel location (x,y).

- Manjunath

Hello,

There is something not very clear for me. If the transformation is linear

HU = Gray_Value * slope + intercept

why, for the same Gray_Value (e.g.white) in some regions HU = 200, and in other regions HU = 800?

Thank you,

Grossu

i can't find slope and intercept in .mhd file

How measure HU value using MIMICS tool with CT images (DICOM format)

ImageJ automatically converts to Hounsfield units.

i saw this PDF https://lhncbc.nlm.nih.gov/system/files/pub9175.pdf and try reproducction with python, but have problem in

I don't have slope and interception y generation with ecuation

Slope = -1/(min-max)

interception = min/(min-max)

new_gray = Gray_Value * slope + intercepción

but when fo example use in values:

print((slope*1000) * 1508 + (intercept*1000))

=999.2236024844719

print((slope*1000) * 221 + (intercept*1000))

=-2.842170943040401e-14

save the imagen and low the contrast but in the paper say that this could better it

please help me

Thank you!

Hi,

When I display a dicom CT slice using 'Mango' viewer, it displays a HU value of roughly 10,000 for a high Z material. For note, the CT has been reconstructed for an extended scale in Siemens Somatom scanner. What details from the DICOM header should I use to calculate the HU in the extended scale? Any suggestions would be much appreciated.

in Python, something like this should do it:

```

import pydicom

dicom_filedata = pydicom.dcmread("input_file.dcm")

corrected_pixeldata = dicom_filedata.pixel_data * dicom_filedata.RescaleSlope + dicom_filedata.RescaleIntercept

```

The direct method by Mark Thurston is a good one. Looks like .pixel_data is now .pixel_array

Another way from the Pydicom docs:

https://pydicom.github.io/pydicom/stable/old/working_with_pixel_data.html#dataset-pixel-array

Modality LUT or Rescale Operation

```

from pydicom.pixel_data_handlers.util import apply_modality_lut fname = get_testdata_file("CT_small.dcm")

ds = dcmread(fname)

arr = ds.pixel_array

hu = apply_modality_lut(arr, ds)

```