A global unimodal thresholding based on probabilistic reference maps for the segmentation of muscle images

Abstract A global probabilistic maps thresholding (PMT) method was applied to characterise intramuscular connective tissue (IMCT) distribution on images of muscle histological sections exhibiting unimodal histograms. Probabilistic reference maps were defined and then used to set-up thresholding rules, derived from linear combinations of parameters calculated from the intensity histogram of the images. This PMT method was objectively compared to Rosin's unimodal thresholding algorithm (RT) and validated by a histochemical quantification of IMCT collagen. Morphometrical parameters of the IMCT (area, length and thickness of the extracted network) were determined for different muscles and used to quantify IMCT distribution differences.