Aurélien Bernard1, Nelly Campolmi, Zhiguo He, Binh Minh Ha Thi, Simone Piselli, Fabien Forest, Jean-Marc Dumollard, Michel Peocʼh, Sophie Acquart, Philippe Gain, Gilles Thuret. 1. *Corneal Graft Biology, Engineering and Imaging Laboratory, Federative Institute of Research in Sciences and Health Engineering, Jean Monnet University, Saint-Etienne, France; †Department of Ophthalmology, University Hospital, Saint-Etienne, France; ‡Department of Pathology, University Hospital, Saint-Etienne, France; §Eye Bank of Saint-Etienne, Auvergne Loire Blood Center, Saint-Etienne, France; and ¶Institut Universitaire de France, Boulevard Saint-Michel, Paris.
Abstract
PURPOSE: A reliable experimental measurement of endothelial cell (EC) viability is paramount in the assessment of new drugs, devices, and surgical processes liable to damage the corneal endothelium, as well as during endothelial bioengineering. We previously used triple Hoechst-Ethidium-Calcein-AM labeling coupled with image analysis to determine the viability and mortality of ECs on the whole cornea, thus defining the new notion of viable EC density. To make it accessible to all, and for improved reproducibility, we have now developed an ImageJ plugin with improved thresholding algorithms. METHODS: The CorneaJ plugin comprised contrast improvement, regional selection of pixels with similar gray levels, simplified thresholding facilitated by a user-friendly images display, and the option of manual touch-up to increase accuracy. After Hoechst-Ethidium-Calcein-AM labeling, the endothelium of 10 human corneas was observed with a fluorescent microscope with motorized stage. The performance of CorneaJ was compared with standard manual thresholding: accuracy was determined by comparison with fully manual selection of viable ECs by an expert; and reproducibility, by calculating the intraclass coefficient and coefficient of variation (100 × SD/mean) of 7 independent observers. RESULTS: CorneaJ was more accurate than the standard thresholding, with a deviation from the expected value of -1.8% [95% confidence interval (CI), -2.7 to -0.9] versus 6.0% (95% CI, 2.8-9.3), respectively, P < 0.001. It was also more reproducible, with an intraclass coefficient of 0.98 (95% CI, 0.954-0.994) versus 0.81 (95% CI, 0.628-0.937) and a mean coefficient of variation of 2.6 (1.4-7.4) and 5.7 (3.4-19.8), P = 0.005. CONCLUSIONS: CorneaJ is a new, fast, and reproducible free image analysis tool that could help standardize experimental measurement of corneal EC viability.
PURPOSE: A reliable experimental measurement of endothelial cell (EC) viability is paramount in the assessment of new drugs, devices, and surgical processes liable to damage the corneal endothelium, as well as during endothelial bioengineering. We previously used triple Hoechst-Ethidium-Calcein-AM labeling coupled with image analysis to determine the viability and mortality of ECs on the whole cornea, thus defining the new notion of viable EC density. To make it accessible to all, and for improved reproducibility, we have now developed an ImageJ plugin with improved thresholding algorithms. METHODS: The CorneaJ plugin comprised contrast improvement, regional selection of pixels with similar gray levels, simplified thresholding facilitated by a user-friendly images display, and the option of manual touch-up to increase accuracy. After Hoechst-Ethidium-Calcein-AM labeling, the endothelium of 10 human corneas was observed with a fluorescent microscope with motorized stage. The performance of CorneaJ was compared with standard manual thresholding: accuracy was determined by comparison with fully manual selection of viable ECs by an expert; and reproducibility, by calculating the intraclass coefficient and coefficient of variation (100 × SD/mean) of 7 independent observers. RESULTS: CorneaJ was more accurate than the standard thresholding, with a deviation from the expected value of -1.8% [95% confidence interval (CI), -2.7 to -0.9] versus 6.0% (95% CI, 2.8-9.3), respectively, P < 0.001. It was also more reproducible, with an intraclass coefficient of 0.98 (95% CI, 0.954-0.994) versus 0.81 (95% CI, 0.628-0.937) and a mean coefficient of variation of 2.6 (1.4-7.4) and 5.7 (3.4-19.8), P = 0.005. CONCLUSIONS: CorneaJ is a new, fast, and reproducible free image analysis tool that could help standardize experimental measurement of corneal EC viability.