Clasificador de células sanguíneas usando procesamiento de imágenes

Abstract

The work focuses on evaluating image processing techniques for blood cell counting and classification. About 25 different techniques were reviewed and two approaches were applied: one based on a custom convolutional neural network to identify and classify cells, and another using a neural network based on YOLO v5 for cell counting. After a comparative analysis, it was determined that the combination of both techniques provided a comprehensive system for accurate counting and effective classification of white blood cells in a single image. A visual interface was created for ease of use and testing with different image banks, demonstrating its utility in medical and diagnostic applications.