Segmentation of Leukocytes in Blood Smeare Images Using Color Processing Mechanism Inspired by the Visual System

Receptive fields of neurons play various roles in biological neural networks. Inspired by receptive fields of rod and cone neurons in the visual system, a neural network model is proposed to segment leukocytes form blood smear images. The network contains three types of cones and a type of rod. Three cones are in response to red, green and blue lights respectively. The rods are in response to brightness. Through a specific receptive field, the network can be trained to segment different colour objects. As leukocyte, cytoplasm and erythrocyte are associated with different colours in blood smear images, the network can be used to well segment them. The experimental results show that it is better for segmentation of large number of leukocytes in an image than other methods.