Classification of Centralized Nuclei from Muscle Fibers
Collaboration
Z. Wu, University of Copenhagen
Research Background
In healthy muscle fibers, the nuclei is positioned at the periphery of the cell; abnormal nuclear positioning, where the nuclei has moved to a more central location, is a common marker for myopathies. In this project, we identify muscle cells and nuclei from fluorescent images of muscle tissue sections. We then classify each nuclei as “attached“ or “detached“ based on its distance from the periphery, in order to count the number of affected cells in the tissue.
Method
We developed a .groovy script in QuPath to automatically process each image. The script performs the following steps:
Identification of Tissue Sample using a trained pixel classifier
Object Detection of nuclei from the DAPI channel, using the built-in QuPath cell detection function; and of cells from the membrane stain, using the ImageJ bridge to run a macro that performs watershed segmentation.
Object Classification of the nuclei by first measuring the signed distance of each nuclei to the nearest cell boundary, and classifying nuclei >5.0um away from the cell boundaries as “detached“. Cells are then classed into those that have “detached” (centrally located) nuclei, and those without any centrally located nuclei.