Quantifying Myotubes from Brightfield Images

I have brightfield images of myotubes and simply want to segment and count the tubes in the image. They are stitched images as of now so the background lighting may vary slightly, but I can crop or create them into stacks if that simplifies the process.

I have tried filtering the images with various morphological filters from MorphoLibJ, adjusting the contrast and brightness, thresholding the images to remove the nuclei and background, and then interactive morphological watershed segmentation. However, the tubes are so close together that I cannot get a clean segmentation.

I was wondering if anyone had any ideas or tips?

I plan on staining the samples to get colored images that will be more easily separated from the nuclei and background, but as of now need the number of myotubes extracted from my current data.

Thanks in advance!