Segmentation of muscle fibres in fluorescence microscopy images

The morphological analysis of muscle biopsy helps in diagnosis of neuromuscular disease. The presence, extent, size, shape, and other morphological appearance of the muscle fibres are important indicators for presence or severity of disease. However, estimation of these parameters by simple visual inspection is inaccurate and subjective and manual delineation of individual muscle fibres from muscle biopsy images is time-consuming and tedious. In this study, two automatic segmentation methods are proposed. Both methods operate on fluorescence microscopy images. The first uses a level set framework and the second one a marker-driven watershed transform. In a first stage, mathematical morphology is used to detect the presence of muscle fibres. The result of this step provides requirements for both segmentation methods (initial contour and markers). Experimental results demonstrate that segmentation of watershed detects fibres contours more accurately and with a lower computational cost.