Phase 3: Refinements

Geodesic reconstruction allows us to use a reliable marker like DAPI to determine which cells overlap an area of marker expression.

Here we make use of the third-party MorphoLibJ library/plugin for morphological operators. MorphoLibJ provides a function for that implements geodesic reconstruction. It's definition looks something like this:

def reconstructByDilation(markerProc, maskProc, connectivity):
  #implentation

We simply create a function to use DAPI as a mask for geodesic reconstruction of a marker image:

def apply_mask(marker, mask):
    """Applies a mask to an input binary using geodesic reconstruction.
    :param marker binary image that with be masked
    :type marker ImagePlus
    :param mask binary image that will be used as a mask
    :type mask ImagePlus
    :return reconstructed image
    """
    ip = GeodesicReconstruction.reconstructByDilation(marker.getProcessor(),
                                                      mask.getProcessor(), 8)
    output_imp = ImagePlus("Reconstructed", ip)
    output_imp.copyAttributes(marker)
    return output_imp

Then we can simply use this function to reconstruct images for each marker showing which DAPI cells express that marker.

# Filter segmented marker images with the DAPI channels that we trust
ki67 = apply_mask(ki67, dapi)
ph3 = apply_mask(ph3, dapi)