White Blood Cells Recognition and Classification using Convolutional Neural Network

Abstract

Effectively recognizing and counting each type of blood cell contributes significantly to the diagnosis and effective management of patients with blood diseases. The majority of hematology laboratories in West Africa do not have equipment dedicated to the automatic classification of blood cells. The integration of AI in Hematology helps to reduce time taken by specialists for the classification of blood cells, to make a rapid diagnosis and to reduce the mortality rate. Following PRISMA model guidelines, this article reviews the main techniques of recognition and classification of blood cells to identify potential approaches that can assist doctors in diagnosis and rapid decision-making. Based on patient dataset from the CNHU-HKM, Benin reference hospital located at Cotonou, the performance of our automatic cell recognition model reached 99% in training, 88% and 90% in validation respectively for images 32*32 and 16*16.