Abstract
The use of machine learning in the medical field is one of the most difficult and thoroughly unsolved problems. Currently, there are many different algorithms for solving problems in the field of diagnostics an segmentation of biomedical images. Researchers are often faced with the challenge of choosing the best method to apply towards their data. We conducted the empirical research and compared 5 algorithms that able to detect anomalies in the medical images: RCNN, Fast-RCNN, Faster-RCNN, Mask R CNN, U-Net, and Residual Neural Network. The advantages of automatic processing of the medical images are apparent: doctors get a convenient software tool that allows them to diagnose the disease faster and reduce possible errors. The task is to study and then select algorithms for further testing on the actual data. The selection and study of algorithms were based on articles describing the architecture and application of computer vision algorithms.