INNOVATIONS IN NEURAL NETWORKS AND THEIR IMAGE RECOGNITION

Abstract

Leukocytes (white blood cells) play a significant role in the processes of blood cell identification and classification. This study considers the application of neural network algorithms for the automatic analysis of blood cell microscopic samples. This method especially focuses on identifying chromatic, geometric, and textural features of leukocytes. An advanced solution is provided for early diagnosing blood Related diseases through image processing and artificial intelligence. Automated systems allow not only to shorten the analysis period but also to reduce the number of errors made by experts. The purpose of this work is to create an efficient neural network algorithm for classification of blood cells samples in order to help in early detection of diseases related to leukocytes. In this study, the efficiency of training models with different architectures of neural networks, namely Recurrent Neural Networks (RNN), ResNet (Residual Network) and Convolutional Neural Networks (CNN), was compared by using different parameters.