Learning based automatic detection of myonuclei in isolated single skeletal muscle fibers using multi-focus image fusion

Accurate and robust detection of myonuclei in single muscle fiber is required to calculate myonuclear domain size. However, this task is challenging because: 1) The myonuclei have a variety of sizes and shapes. 2) Imaging techniques exhibit myonuclei that are often overlapping. 3) Inhomogeneous intensity due to DAPI concentration in heterochromatin, abundant in mouse nuclei, results in a speckled appearance inside each myonucleus. In this paper, we propose a novel automatic approach to robustly detect the myonuclei in isolated single skeletal muscle fibers. The original z-stack images are first fused into one all-in-focus image. A sufficient number of ellipse fitting hypotheses are then generated using the myonuclei contour segments. A set of morphological features are calculated from the ellipses and utilized to train a support vector machine (SVM) classifier to choose the best candidates. A modified inner geodesic distance based clustering algorithm is used to produce the final results. The proposed method was extensively tested using 42 sets of z-stack images containing about 1500 myonuclei. The method demonstrates excellent results outperforming current state-of-the-arts.