AN APPROACH FOR SEGMENTATION OF SKELETAL MUSCLE FIBERS IN HISTOLOGICAL IMAGES
Marlon Polo de Melo, Joaquim Cezar Felipe · 2012
In order to study and identify the structure and function of healthy/diseased human skeletal muscle tissue, some measurements on muscle fibers, such as fiber size, minimum diameter and fiber type composition, are key data to evaluate the quality of muscle. Estimation of such measures by simple inspection is inaccurate, subjective, and time-consuming. Thus, in this paper we propose an approach for automatic segmentation of images from histological sections of skeletal muscle fibers with histochemical reaction of the enzyme adenosine triphosphatase. We have evaluated three approaches for automatic thresholding, namely, Multilevel Otsu's method, Watershed operation and Gaussian mixture model. In addition, aiming to define clusters of fibers, we have evaluated three methods, namely, Voronoi diagram, Kumar, and Watershed, complemented by some morphological operations. Our initial results indicate that this study can help to analyze the muscle tissue by an accurately and quickly way, when compared with manual analysis performed by experts.