The use of hematopoietic progenitors in toxicology often requires scoring colonies, micro- and macroclusters, and evaluation of their distribution while taking into account their shape. Unfortunately, this evaluation is long and tedious, and classification errors may occur. Image analysis can improve the method if there is an automatic way of performing the classification analysis. It has been previously demonstrated that cell conglomerates can be automatically detected. The present paper describes three image analysis algorithms and reports on their capabilities to analyze cell heterogeneity in culture. The comparative tests are made on computer-generated images representing different cell distributions. Combining two of these three algorithms gives fairly good results on complex images created by adding various simple distributions.