Hybrid Image Processing Technique for the Robust Identification of Unstained Cells in Bright-Field Microscope Images

Abstract

The initial results of a robust technique for the identification of unstained cells in bright-field microscope images captured with high magnification lenses are presented. This is part of an extensive study on the effects of DNA damage to cell viability induced by multi-photon absorption. The method makes use of a variety of image processing algorithms together with expert knowledge of cell morphology during the cell cycle. Single stem cells are identified by first detecting their cell nucleoli and then clustering into nuclear groups. Experimental results, for a representative sample of images, are compared with ground truth labeling by skilled biologists. Despite the low contrast and high variability in appearance of cells in bright-field microscope images, the reported technique displays a detection rate of over 79% for the correct number of identified cells. The methodology implemented has sufficient accuracy and speed for the development of high- throughput robotic systems.