DICOMwaterequivalent

This is a python 3 script / program to calculate the patient's water equivalent area (A_w), water equivalent circle diameter (D_w), and area-equivalent circle diameter, from 16 bit CT DICOM images. A_w and D_w consider tissue attenuation for calculating the patient size for size-specific dose estimates (SSDE) in CT, as proposed by AAPM Task Groups 204 and 220.

It can be used as a Python function or as a standalone Python script. It returns the following information:

• ROI area
  area of the automatically detected ROI, in mm².
• ROI-equivalent-circle diameter
  diameter of the circle with the same area as the ROI, in mm.
• water-equivalent area (A_w)
  ROI area normalised by its density to the density of water, in mm².
• water-equivalent-circle diameter (D_w)
  diameter of the circle with the same area as A_w, in mm.
• hull area
  area of the convex hull around the ROI, in mm².
• hull-equivalent-circle diameter
  diameter of the circle with the same area as the hull, in mm.
• image
  the input DICOM file, overlaid with contours of the ROI and hull area, and all above measurements.

The ROI is determined as the largest region of tissue above a manually chosen density threshold (in HU) and all tissue enclosed by it. Thresholds around -100 to -300 will usually separate the patient from the examination table well, without excluding any patient tissue. If no ROI can be detected, it will return a ROI of 0mm².

⚠️ Always check the output image for correct ROI placement. The ROI is automatically placed around the largest contour with HUs above the ROI HU threshold. You must manually set the ROI HU threshold. Confirm that the patient contour is inside the displayed ROI outline, and that the ROI does not include the CT examination table, clothing, implants, ECG leads etc. For measurements other than A_w and D_w, also avoid including air. Exclusion of implants is not (yet) possible with this script.

⚠️ This software and its results should not be used blindly without adequate professional judgment. Verify that the pixel-to-mm scaling is done correctly by checking the ROI area with the CT manufacturer's software. Read LICENSE for further disclaimers.

Contents

• Requirements
• Standalone
• Python function
• More information
  • Using SSDE factors
  • Sources / suggested reading
  • Contact

Requirements

Modules cv2 and pydicom:

$ pip3 install opencv-python pydicom

Download the file DICOMwaterequivalent.py and place it in your working directory. Alternatively, you can put it in one of the other directories that Python checks for modules. To list these directories: $ python3 -c 'import sys; print(sys.path)'

On Linux and MacOS, you might need to make it executable:

$ chmod +x DICOMwaterequivalent.py

Standalone

You can call DICOMwaterequivalent.py from the command line:

Input

$ ./DICOMwaterequivalent.py <filename> <threshold> <ww> <wl>

• filename: DICOM file (absolute or relative path).
• threshold: ROI contour threshold level (in HU).
• ww: (optional) window width (default: 1600).
• wl: (optional) window level (default: -400).

You need to specify both ww and wl, or neither.

Output

• Console: output values (float)
• GUI: image with ROI and hull contours, overlaid with output values (int)

Example

$ ./DICOMwaterequivalent.py 480.0.dcm -250
roiArea: 21445.37160223951
roiEquivalentCircleDiameter: 165.24253440174482
waterEquivalentArea: 26328.33432149283
Aw: 26328.33432149283
waterEquivalentCircleDiameter: 183.0908965654292
Dw: 183.0908965654292
hullArea: 23966.764703619567
hullEquivalentCircleDiameter: 174.68666972614523
Press any key in the GUI to quit

* Example dicom file courtesy of Patient Contributed Image Repository patient 54879843, available from http://www.pcir.org/researchers/downloads_available.html

Python function

You can call DICOMwaterequivalent() from your own python script:

Input

$ python3
>>> from DICOMwaterequivalent import DICOMwaterequivalent
>>> DICOMwaterequivalent(filename, threshold, window)

• filename: DICOM file (absolute or relative path).
• threshold: ROI contour threshold level (in HU).
• window: (optional) view window for output image, as tuple (ww,wl). If omitted, no image will be outputted.

Returns

Dictionary containing numerical values as float, image as numpy array.

Example

$ python3
>>> from DICOMwaterequivalent import DICOMwaterequivalent
>>> equiv = DICOMwaterequivalent('480.0.dcm', -250, (1000,40))
>>> print(equiv)
{'roiArea': 21445.37160223951, 'roiEquivalentCircleDiameter': 165.24253440174482, 'waterEquivalentArea': 26328.33432149283, 'Aw': 26328.33432149283, 'waterEquivalentCircleDiameter': 183.0908965654292, 'Dw': 183.0908965654292, 'hullArea': 23966.764703619567, 'hullEquivalentCircleDiameter': 174.68666972614523, 'image': array([[[0, 0, 0], ... ]]], dtype=uint8)}
>>> import cv2
>>> cv2.imwrite('out.png', equiv['image'])

More information

Using SSDE factors

SSDE conversion factors can be calculated from D_w depending on the phantom used for CTDI_vol estimation. The following formulae are for D_w in mm:

For 32 cm phantoms:

conversion factor = 3.704369 * e^(-0.003671937*Dw)

For 16 cm phantoms:

conversion factor = 1.874799 * e^(-0.003871313*Dw)

(Derived from the formulae with D_w in cm in AAPM Report 204. Note that Report 220 Appendix A clarifies that D_w should be used, which is reported as 'effective diameter' in Report 204)

The resulting conversion factors can then be applied to phantom-based dose estimates (Gy*cm or Sv), to achieve a size-specific dose estimate. Of course it makes most sense to do this for every slice of a scan, not just a single slice.

Sources / suggested reading

AAPM report 220: McCollough C, Bakalyar DM, Bostani M, Brady S, Boedeker K, Boone JM, Chen-Mayer HH, Christianson OI, Leng S, Li B, McNitt-Gray MF. Use of water equivalent diameter for calculating patient size and size-specific dose estimates (SSDE) in CT: The Report of AAPM Task Group 220. AAPM report. 2014 Sep;2014:6.

AAPM report 204: Boone JM, Strauss KJ, Cody DD, McCollough CH, McNitt‐Gray MF, Toth TL. Size‐Specific Dose Estimates (SSDE) in Pediatric and Adult Body CT Examinations: Report of AAPM Task Group 204 .... College Park, MD: American Association of Physicists in Medicine; 2011.

Cheng PM. Automated estimation of abdominal effective diameter for body size normalization of CT dose. Journal of digital imaging. 2013 Jun 1;26(3):406-11.

Contact

Any questions or comments? Mail me: caspar@verhey.net

If you run into trouble, please mention:

• OS;
• python version;
• whether you have the required libraries installed;
• your command line arguments or python code (screenshots preferred);
• if possible: one of the DICOM files you're trying to use.