This talk presents two applications of machine learning to classification problems in cell microscopy. The main purpose is to demonstrate the high degree of flexibility and reliability of this technology. The first part of the talk concerns the detection of leukemia in blood smears. This is done by classifying nucleus, nucleoli and cytoplasm, followed by higher-level morphological analysis. The second part of the talk is dedicated to classification in live-cell bright field microscopy, which complements information obtained with fluorescence microscopy for the purpose of tracking the cell cycle and for phenotyping.