Hi Bronwyn,

could you please post an image here so I can see your problem?

Kind regards,

Jason

This should work.

1. Process, Filters, Variance. I radius of 2 or 3 works for me.

2. Threshold with Triangle algorithm.

3. Analyze particles. Set minimum size to avoid very small objects.

4. Measure the area of the ROIs. This will give you area occupied by cells. Then all you need is the area of the total image and you can get % confluency.

Thanks Christopher B O'Connell I will give that a try.

Hi Ioannis Jason Limnios , it isn't a fabulous image but they're difficult to image with being so flat! I know that to stain would be a lot easier but Im trying to assess confluencey over time so cannot take for staining.

Here is what I got. I had to slightly modify what I described earlier. I cropped out the scale bar and converted to an 8 bit greyscale image. Then I used a variance filter with a radius of 4 pixels. Thresholding was done with Otsu algorithm. Process, binary, fill holes to close up some of the gaps in the binary. Analyzed particles with minimum of 200 pixels2. You need to use spatially calibrated images to measure the area of ROIs and total image. Your image was not calibrated.

Overall, not bad. The huge flat cells have low variance in intensity at the periphery so they didn't get picked up totally, but its good enough you will get a pretty accurate result.

Christopher B O'Connell That is absolutely brilliant, thank you so much!!

Bronwyn K Irving,

Christopher B O'Connell's method looks great for a quick and simple analysis and any more effort on those images would give diminishing returns.

Also, the EVOS 4x is pretty bad for seeing contrast and detail. You could improve your images for analysis by taking more 10x images and automating the process with a macro.

Since it is also a fluorescent microscope, you might consider using a constitutive cytoplasmic GFP line (by lenti), especially if you are analysing many images and need higher accuracy.

I have attached old images of primary MSCs 24 hrs after lentiviral transduction, no selection. This analysis was done in ImageJ and can be automated with a macro to churn through images. I'd improve it by allowing the cells more time to express GFP, followed by FACS sorting and slightly overexposing.

Kind regards,

Jason