Recognition of Leukocytes on Smears of Peripheral Blood and Bone Marrow Using a Neural Network Approach
Yu. V. Zorin, M. A. Avanesov, А. Н. Проничев, A. D. Palladina · Physics of Atomic Nuclei · 2022
Abstract The article studies the problem of classifying leukocytes in images of peripheral blood and bone marrow preparations with multiple contact of leukocytes with each other for automated diagnosis of diseases of the hematopoiesis system. The proposed approach is based on the definition of a class of leukocytes by a combination of the K-means method and a convolutional neural network. The application of the K-means method is preceded by the implementation of the watershed algorithm with distance conversion. On the basis of the results of the experiment, the accuracy of recognition of lymphoblasts, granulocytes, monocytes, and lymphocytes is evaluated. The proposed solutions can be further applied in decision support systems in the diagnosis of acute leukemia.