Signal (dot) counting in fluorescence in-situ hybridization (FISH) images that relies on an automatic focusing method for obtaining clearly defined images is a time-consuming procedure prone to errors. Our recently developed system has dispensed with automatic focusing, and instead relies on a neural network classifying focused and unfocused signals into valid and artefact data, respectively, and thereby discriminating between in- and out-of-focus images. However, to train the classifier accurate labelling of the image signals is required. GELFISH is a Graphical Environment for Labelling FISH images that enables the rejection of unanalysable nuclei and labelling of FISH signals simply and rapidly. GELFISH is flexible and can be modified easily for additional FISH applications. Also, implemented using popular software, the environment can be employed on any computer by any user. Finally, GELFISH is proposed in controlling a classifier-based dot counting system.