Gradient-based cell localization for automated stem cell counting in non-fluorescent images

Counting the numbers of cells in microscopic images is an important task to measure the level of cell proliferation quantitatively. Existing approaches require either fluorescing cells or manual counting, both of which are expensive and time-consuming. In this paper, we introduce a novel algorithm for localizing cells in non-fluorescent gray-scale images. Using the fact that the regions around cells undergo drastic intensity changes, we apply the image gradient filters to single out the cell regions that exhibit strong filter responses. The detected spatially high-frequency regions are further post-processed to accurately crop out tight cell bounding boxes and remove noise signals. On the microscopic stem cell images, we demonstrate that the proposed approach can yield cell counting accuracy comparable to manual counting, while fast enough to handle about 10 images per minute in MATLAB on a standard machine. © 2014 The Korean Tissue Engineering and Regenerative Medicine Society and Springer Science+Business Media Dordrecht.