Automatic Cell Viability Determination In Bright-Field Microscope Images

Abstract

Quantification of non-viable cells in bright-field microscope images allows for an in vitro assessment of the viability of cultured cells. The identification and quantification of non-viable cells in a cell culture is conventionally achieved with viability stains. A disadvantage of viability stains is that they might be invasive and even toxic to a cell culture. For this purpose, a non-invasive keypoint-based cell viability determination method without the use of viability stains, has been developed. This method formulates unstained non-viable cell detection as a supervised, binary pattern recognition problem and presents a set of features that are suitable for distinguishing between unstained viable and non-viable cells in bright-field micrographs. Experimental results for a representative sample of micrographs are compared with the ground truth non-viable cell count obtained from fluorescent micrographs. Despite the low contrast and high variability in appearance of cells in bright-field images, the method yield a classification rate in excess of 88% for non-viable versus viable-suspended cells and of more than 65% for nonviable versus viable-adherent cells . Thus, the developed method has been proven to be a feasible alternative for cell viability determination in bright-field micrographs.