Bug fix in ROIContour.getBinaryMask(): up sampling contours

General
1

When getting contour mask from mesh, if the image z spacing (dZ_s) is smaller than the minimum mesh vertices along the Z axis (dZ_m), the mask is generated at every dZ_m/dZ_s slices. For example, on some CBCT images, when going from resolution (1 x 1 x 3) mm^3 to (1 x 1 x 1) mm^3, the mask is generated every 3 slices (figure below).

Pasted image 20251011184851
Pasted image 20251011184851954×584 109 KB

Here is the cause of the problem
file path:

/opentps_core/opentps/core/data/_roiContour.py:ROIContour.getBinaryMask()
line 119


        if len(zDiff) == 0:

            from opentps.core.data.images._roiMask import ROIMask
            return ROIMask(imageArray=None, name=self.name, origin=contourOrigin, spacing=contourSpacing,
                           displayColor=self._displayColor)
        else:
            if np.isfinite(zDiff[0]):
                 contourSpacing[2] = zDiff[0]
            else:
                 contourSpacing[2] = minSpatialResolution

The fix is very small:

        if len(zDiff) == 0:

            from opentps.core.data.images._roiMask import ROIMask
            return ROIMask(imageArray=None, name=self.name, origin=contourOrigin, spacing=contourSpacing,
                           displayColor=self._displayColor)
        else:
            contourSpacing[2] = zDiff.min()
            # if np.isfinite(zDiff[0]):
            #     # contourSpacing[2] = zDiff[0] 
            # else:
            #     # contourSpacing[2] =  minSpatialResolution

note that numpy.min() function automatically ignores np.nan entries.

2

Hello Hossein,

I’ll be sure to add this fix also in the MR (here is the link so that you can follow our progress).

Thank you for the fix !

1 Like