Reading in an External cell boundary

Dear CellProfiler-team,

is there a way to have CellProfiler read in a file and use it as a boundary for the SecondaryObjectIdentification?

With most of my DIC-images a reliable cell segmentation can only be done semi-manually, I fear. So with a bit of help of ImageJ, I end up with a composite image (or many single channel images) as given in the attachement (unfortunately, I can only upload one image) where the fat yellow line is supposed to be the cell boundary. Supplying a DAPI-channel for primary object identification (nuclei) works like a charm. But I am now having difficulties telling CellProfiler to expand the nuclei until they either meet each other or meet a yellow boundary. So far, CellProfiler either only recognizes the yellow lines themselves as secondary objects or it clumbs up nearly all cells into one giant secondary object. In no case the boundary is recognized as a boundary.

Is there a way that allows CellProfiler to accept the boundary image as containing boundaries?

A solution as offered in the forum (viewtopic.php?f=14&t=4127#p12422) fails mostly due to my low overall fluorescence signal from the cytosol, even if I sum up all channels.


I’m attaching a pipeline that should get you closer to your goal. Since I just have the one image, I had to do some pre-processing to identify the nuclei and the outlines. It works for this image, but ideally, the outlines should be a separate file loaded and matched with the corresponding cell image.

In any case, the basic strategy was to identify the outlines and then invert the image so that the cells would be light, separated by the black outlines. Once the nuclei are identified (by whatever means), IdentifySecondary works fine on such an image, esp. if the threshold is set manually.

assay.cppipe (11.3 KB)