Figure1. Image capture and processing for the samples under complicated backgrounds.
Cells in suspensions were deposited into a hemocytometer, and images of complex background was taken and cropped to 1x1 mm. Representative images of one area with 16 smallest counting chambers are shown. Panel A shows the original RGB image captured by microscope. Panel B, the image was converted to 8-bit and resized. After compression, machine learning-based ilastik was used to distinguish the background from yeast cells. Panel C shows the process that a user-defined class label was attached to the images with complex background. Whereafter, ImageJ macro was used to optimize the batch of images. Black-and-white images were presented first. Panel D shows the operation to fill the gap with the function of ImageJ, which are marked by the red circles. Panel E shows merging cells split by a single pixel line via the “Watershed” function, which are marked by red clipper. Area can be used to assess the objectives in images with ImageJ tool. The Area command was applied in panel F via the “Analyse Particles” function. After setting the threshold in the Analyse Particles command, cells counted automatically are highlighted and numbered in an overlay on the image as indicated in panel G.