Application of Deep Learning Methods for Automation of Leukocyte Microscopy

Hematology is data-rich discipline and at the same time the hematological data are complex and often difficult to understand. Using deep learning methods, it is possible to solve a range of problems related to analyzing a large number of hematological preparations. This is particularly relevant for such activities as mass medical examination and distance microscopy as part of telemedicine system. This article discusses the possibility of using deep learning models based on convolutional neural networks to automate leukocyte microscopy. An approach to solving the problem of detecting and classifying cells in microscope images is proposed. The key components of biomedical image analysis pipeline are implemented including data preprocessing module, segmentation and image classification neural network models. The proposed pipeline is evaluated, and further directions of development are considered according to the obtained performance characteristics.