I’m counting the number of neuronal cells from a set of microscopic images by segmenting the regions with high pixel intensity overlap across two channels (One for nuclei with Dapi stains and one for neurons with Tuj1 stains). I used IdentifySecondaryObject for this purpose.
Some of the nuclei were for neuronal progenitor cells so there was no one-to-one correspondence between neurons and nuclei (there were more nuclei than neurons).
To remove noises that can interfere with the counting performance (especially the network of neurites) from the image, I restricted the image for the neuronal channel to areas around nuclei by objects masks made from dilating the nuclei objects segmented by IdentifyPrimaryObject by 3 pixels. I then used the Threshold module to phase out parts of the restricted neuronal image with dim signals. I then called IdentifySecondaryObject with the nuclei masks as the reference (since they will be contained in the restricted neuronal cells if there is a cell), but unexpectedly, some areas of the neuronal image with no pixel intensity were also counted secondary objects (See image below). They were the false positives of the counts and happened with virtually every strategy (watershed, gradient, distance etc.). Can someone help me find out why this would happen and can this be stopped by adjusting the parameters of or using some other options from the module?