Automated imaging, identification, and counting of similar cells from digital hologram reconstructions
This paper presents our method, which simultaneously combines automatic imaging, identification, and counting with the acquisition of morphological information for at least 1000 blood cells from several three-dimensional images of the same sample. We started with seeking parameters to differentiate between red blood cells that are similar but different with respect to their development stage, i.e., mature or immature. We highlight that these cells have different diffractive patterns with complementary central intensity distribution in a given plane along the propagation axis. We use the Fresnel approximation to simulate propagation through cells modeled as spheroid-shaped phase objects and to find the cell property that has the dominant influence on this behavior. Starting with images obtained in the reconstruction step of the digital holographic microscopy technique, we developed a code for automated simultaneous individual cell image separation, identification, and counting, even when the cells are partially overlapped on a slide, and accurate measuring of their morphological features. To find the centroids of each cell, we propose a method based on analytical functions applied at threshold intervals. Our procedure separates the mature from the immature red blood cells and from the white blood cells through a decision based on gradient and radius values.