Application of machine learning algorithms in morphopathology and in assisted reproductive technologies

Machine learning models are used everywhere to analyze images, signals, and videos. At first glance, this is a well-designed process that involves the stages of data collection, mark-up, and training a model, and, as a result, its application in a particular field (recognition of vehicle plate numbers, smartphone faces, etc.). However, everything is much more complicated in the field of medicine: the use of artificial intelligence models is a serious challenge. Machine learning methods are becoming more and more used in morphological sciences and biomedical studies. The introduction of artificial intelligence for image analysis can lower the burden on an operator (a pathologist, a histologist), eliminate the factor of subjective assessment, and reduce the likelihood of an error. This review provides a brief excursion into the history of machine learning methods, considers the examples of their use in two areas where they are most widespread: morphopathology and assisted reproductive technologies, and also indicates the limitations and difficulties that developers face when training neural networks. Conclusion: The authors also propose solutions to overcome the difficulties associated with the collection and joint marking of data, and model training: creation of a high-quality infrastructure, attraction of highly qualified specialists who mark data, an advanced scientific approach to artificial intelligence technologies; cloud platforms are offered to be used as a basis for the scalable storage and analysis of biomedical data. © 2021, Bionika Media Ltd.. All rights reserved.