• yeast;
  • viability;
  • microscopy;
  • image analysis;
  • artificial neural network


A dedicated microscopy imaging system including automated positioning, focusing, image acquisition, and image analysis was developed to characterize a yeast population with regard to cell morphology. This method was used to monitor a stress-model alcoholic fermentation with Saccharomyces cerevisiae. Combination of dark field and epifluorescence microscopy after propidium iodide staining for membrane integrity showed that cell death went along with important changes in cell morphology, with a cell shrinking, the onset of inhomogeneities in the cytoplasm, and a detachment of the plasma membrane from the cell wall. These modifications were significant enough to enable a trained human operator to make the difference between dead and viable cells. Accordingly, a multivariate data analysis using an artificial neural network was achieved to build a predictive model to infer viability at single-cell level automatically from microscopy images without any staining. Applying this method to in situ microscope images could help to detect abnormal situations during a fermentation course and to prevent cell death by applying adapted corrective actions. © 2011 American Institute of Chemical Engineers Biotechnol. Prog., 2011