Automated dicentric scoring system in Singapore nuclear research and safety initiative

The dicentric chromosome assay (DCA) serves as the gold standard for quantifying ionising radiation exposure in individuals. However, the conventional manual scoring method for DCA is both time-consuming and mentally demanding. Consequently, numerous research groups and commercial entities are actively pursuing the integration of artificial intelligence to automate this process. In this study, we present methodologies and techniques employed at the Singapore Nuclear Research and Safety Initiative (SNRSI) for the development and optimisation of an in-house automated dicentric chromosome scoring tool. Our approach involves the utilisation of thresholding and watershed methods to identify chromosomes within a metaphase. Subsequently, these identified chromosomes are fed into a trained convolutional neural network (CNN) for classification and centromere number assignment. The cumulative centromere count is then calculated to determine the acceptance or rejection of the metaphase for scoring. This integration of advanced image processing techniques and machine learning algorithms would streamline and enhance the efficiency of the dicentric chromosome scoring at SNRSI.