Biotechnology is still a growing branch of industry with an estimated world market of US$ 43.7 billion in 2001. Supported by a large number of ongoing genome projects, the ultimate goal is to identify proteins or other metabolites which can be used to cure diseases.
With the help of Biotechnology it is possible to produce such proteins by genetic modified bacteria and mammalian cells. In such fermentations the biomass concentration of the organisms is the main indicator for the cell growth, and with it the productivity. Another indicator during fermentation is the cell viability (the ratio of dead and vital cells). In most cases the cell viability is estimated by microscopy and manually staining of the cells which is time consuming and imprecise. There are also particle counter available. They can be used to estimate the cell concentration by electric resistance, but lack the ability to differentiate between dead and vital cells. Since 1998 a cell counter with automatic staining, dilution and cell counting of dead and vital cells is available, which is an advantage over the other methods. But in all of these methods a sample has to be taken out of the fermenter, holding the risk of infection.
We introduce an image analysis system which works together with an in-situ probe to count the cells and differentiate between dead and vital cells inside the fermenter. The probe is composed of a digital camera, darkfield lighting, lens and a movable cell chamber for counting the cells. The image analysis system works on the basis of darkfield microscopy images. The method of darkfield microscopy permits the differentiation of dead and vital cells by their different refraction indices, showing the vital cells with a bright cell border on the black background and a black inside (except cell organelles). In contrast, dead cells show a very bright border and also a very bright inside. The system is used to estimate the amount of yeast cells. Saccharomyces cerevisiae belongs to the budding cells, so the images include single cells, budding cells and cell clusters. These different cell formations have to be detected as well as cell size distribution which gives also important information about the fermentation process.
The regions are detected by a dynamic threshold, which compensates the different light adjustments of the microscopy. Features of the regions like size, excentricity, form factor, mean and variance of the region intensity are extracted and cell volumes calculated. By a polynomial classifier the regions are detected as vital or dead single and double cells or cell clusters. In other cell counting systems the amount of cells in cell clusters is estimated by division of region size and average cell size only. In our system each cell and its size in a cell cluster can be detected by using active contours. The contours are initialized randomly in the cluster region. Internal, external and image forces pulls the contour to the maximal intensity, the border of a cell, and detects each cell. With this tool an automatic analysis of undiluted cell suspension with high cell concentration is possible. The technique is also applicable for other eukaryotes like hybridoma cells.