Conformity Index Calculation: to Or or not to Or Body Mask and PTV(ROI)

General
1

Hi there,
When calculating conformity index (CI), I realized some of my data have CI>1.

This happened because the body mask and the PTV mask were none overlapping when the data was exported from the clinical TPS (picture below).

The fix would be to do an “OR” operation between the body mask and the PTV (code after the picture). However, I am not sure that the assumption PTV always being inside the body would apply in all cases? Alternatively, we could leave it to the user to ensure that the body mask is overlapping with all the structure masks inside the body.

image
image1311×826 174 KB

Here’s the quick fix:
File:
/OpenTPS-brachyutils/opentps_core/opentps/core/data/_dvh.py
Class:
DVH.conformityIndex()
line 393 to 402

assert self._prescription is not None
percentile = 0.95  # ICRU reference isodose

body_mask = body_contour.getBinaryMask(
    self._doseImage.origin,
    self._doseImage.gridSize,
    self._doseImage.spacing).imageArray

body_mask = np.ma.mask_or(body_mask, self._roiMask.imageArray)  # include target volume in body mask

# prescription isodose volume
isodose_prescription_volume = np.sum(
    self._doseImage.imageArray[body_mask] >= percentile
    * self._prescription) #V_RI
2

Hi Hossein,

You are indeed right, if the two mask does not overlap, CI should be computed on the union of the two.
I added a warning in case the user did not intended for that. I pushed the fix to the merge request.
My colleague was about to merge the Brachy Utils patch so we are just in time !

Have a nice day !
Eliot

1 Like