Maintaining stem cell lines requires selecting colonies for passage based on characteristics that are visible under the microscope. Currently, stem cell colony selection is performed manually based on visual inspection of phase contrast images by trained biologists. The colony selection process is difficult, and its consistency and repeatability are of concern.The expert biologists in this study agreed on a set of rules governing their visual inspection and stem cell colony ranking which improves the overall selection consistency. We describe an approach to increasing consistency of ranking based on an automated generation of quantitative information about the attributes that make up the experts’ rules. This talk will describe the design of an automated system for scoring stem cell colony images and our plan to quantify all sources of error in the process, from creating the biological rules to translating them into image features and using them.