Detección de cáncer en la sangre mediante el procesamiento de imágenes aplicando redes neuronales convolucionales

Abstract

Abstract: Leukemia is one of the most lethal diseases for humans, being a type of cancer that affects blood cells. In 2024, the American Cancer Society in the United States estimated that around 62,770 new cases will be diagnosed, and approximately 23,670 deaths are expected to occur due to this disease. The diagnosis of leukemia is a complex process for healthcare professionals, as determining the type and stage of the disease also poses challenges. The diagnosis is made by taking into acount the changes in the white blood cells, which are divided into five types: neutrophils, eosinophils, basophils, monocytes and lymphocytes. This thesis project proposal focuses on creating an ecosystem that includes gathering information (a database of blood smear images), developing a convolutional neural network for classifying peripheral blood images using architectures such as ResNet, VGG, and EfficientNet, and creating a prototype that will allow healthcare users to easily and intuitively access the necessary functionality to generate classifications and, most importantly, validate the results obtained. We will use methodologies already established and proven within the execution of Artificial Intelligence projects both for the systematic analysis of the literature, for the creation, training and validation of AI models, as well as for the construction of the validation prototype, this guaranteed a correct and efficient workflow integrating flexibility and adaptability between the defined phases.