Training of automatic preclassification during operation – the way to create an autonomous automatic analyzer of complex morphology
The development of a microscopy combine designed to automate the analysis of complex morphological objects is presented. Automatic tecniques of combine analysis form the results of the “preclassification” class with the control of automatic results by the user. The harvester has means of automatic adaptation to the current slide and to the population of objects of analysis presented in the streams of slides served by the team of combines of laboratories. Local adaptation optimizes the quality and speed of the slide scanning. Adaptation to the population of objects of analysis is carried out by training neural networks of combine analyzers using a common database of automatic preclassification adjustments by qualified laboratory users. Training is used to increase the accuracy of preclassification with the ultimate goal of creating a stand-alone analyzer without user control.