Segmentation of whole blood and size of cells


I am trying to do segmentation of multi-layered whole blood in a cuvette with a thickness of around 30um. The blood should not be diluted and smeared as it is seen in the literature.
I am trying to do automated segmentation and find the size of the cells. It is not necessary to find all blood cells but only the clearest in the picture. I have read about different techniques such as watershed, Hough transform, Shape analysis etc, but all are performed on smeared images. I have attached an example image. The images are obtained from microscopy with a digital camera.

Thanks for your help,


how many cells do you count by hand in the image you posted?

1 Like

Actually I am not interested in finding all the cells, but maybe use circularity or something else to point out e.g. 5-10 clear cells and find the diameter of those cells.

We are having trouble to get i thinner cuvette, but we are also thinking about using a Selective Plane Illumination Microscopy to get a thin plane, where it would be easier to image cells.

yes, honestly I think you have to work more on image acquisition and/or sample preparation than on the image analysis step.
At the moment, I think it will be hard get reliable results on cells segmentation.

1 Like

yeah, that is also my thoughts. Do you have any knowledge about other people trying to do analysis on whole blood instead of smeared/single plane images?

Think that SPIM could be the solution. The problem is that is it not possible to do any sample preperation with the intended purpose. Just straight from blood sample -> optical imaging system -> images -> Find size of cells.