The algorithm identifies and counts cells using machine learning approach. It detects cells with different staining types (IHC, multiplexing, chromogenic and fluorescence) from learning images. Cell Recognition is able to identify five different types of cells based upon color, edge and texture criteria.
The stages of image analysis with the Cell Recognition algorithm are as follows:
Color plan selection
Using color-deconvolution tools, the adapted color plan is selected. It can be the RGB image or only specific layer like DAB or IF channels.
Creation of a learning model
Different kind of cells are selected by the user to define a training data set. Based on this selection, the RFT algorithm creates its classification model and establishes a probability map for each type of cells.
Filter based on probability threshold
Probability thresholds are determined to define the appropriate class to each segmented cell. Those thresholds are adapted based on tissue type.
The results show the number of cells by class and measure the number of neighbours for each cell. A cell description is available.