So I am new to Cell Profiler and I am trying to measure the amount of IBA1+ cells (stained green) from mice brain tissue. I created a pipeline with “gray to color” to get the image in the green channel and “identify primary objects” to count the cells, but I keep getting inconsistent counts with the different thresholding methods. I found otsu to be good (2 class), but does anyone have any suggestions on how to make my pipeline better? I tried adjusting the correction factor for Otsu but it doesn’t seem to help much. I can’t share a photo but I’m using a merged image of different stains.
Firstly, I am wondering if your changing your grey scale image to color & using for thresholding. If so, it might be appropriate since this module is meant to create a color image for ex. composite image. For thresholding you chose the appropriate channel image.
Secondly, different thresholding methods giving different results is usual since the algorithm being used is different. You can always chose the thresholding that suits for your image.
Actually, it is very hard for us to suggest a method without looking at the image.
If it is a basic cell segmentation & IBA1 marker staining (i.e. 2 channels), you could refer any of the basic pipelines in the example page. If you are new to CP, you could watch the workshop here & download quick start guide here.
Without looking at a sample image or a pipeline that you are trying, it would be difficult suggest improvements for your pipeline. So, please do share it if possible.
Which version are you using?
CON (L) ACC.tif (779.4 KB)
Hi @lakshmi Lakshmi, thanks for the response. Attached is an examplar image. My issue is that when I try to quantify the cell count in different regions of the brain using the same thresholding strategy, it seems inconsistent when compared to my hand count of the cells. I am using version 4.0.
Thanks. I checked you image though I am not sure if you use DAPI or other channel (red signal in your RGB image) to count your cells. In either case it shouldn’t be a problem as along the signal is almost even in the image. This could also be achieved by appropriate filtering method. Nevertheless, if you are using two different segmentation approach for the same image because of the uneven signal, you could use the same module twice in the pipeline with the different settings. Later you can combine the objects from both the steps using “CombineObject” module as described in the later part of the video here.
If this is not solving your problem, you may have to share your pipeline that you are trying so that we could help you better.