One of the ways to achieve this is by using connected component analysis or hough circles, but I believe you may need to use a better contrast image in the first place.
If there is a better contrast between the cell boundaries and the cell body, you can first extract the cell boundaries by thresholding. This provides you with a binary image where white region (or 1s) represents the cell boundaries. If you invert this binary image, now the cell body becomes the white part and the boundaries become black. To avoid connectivity between the cell body and the close neighboring cell bodies, you can erode the white regions by using the imerode function.
If everything is proper, now each cell body will an independent white region in the image. you can use regionprops function to find out the individual area.