Adaptive 3D cell segmentation and tracing algorithm using convex separation and histogram information for vivo images

In biology and medicine, it is critical to study the amount, distribution, movement, status, and behavior of cells. To do this, one should well segment cells from three dimensional (3D) cell images in prior and precisely extract the moving path of each cell. In this work, an advanced 3D cell segmentation and tracing algorithm is proposed. It applies local histograms to perform adaptive thresholding. Moreover, based on the observation that a cell is always nearly convex, the technique of shortest path segmentation is applied to disconnect the over-merged cells. Furthermore, the overlapping cell areas between two adjacent layers are adopted to refine segmentation results. Then, the moving path of a cell can be traced by determining the cell with the shortest distance and the most similar characteristics in the next 3D image. With the proposed algorithm, 3D cell images can be accurately segmented and each cell can be well traced.