Uso de AlexNet como red convolucional para la clasificación de imágenes médicas en el diagnóstico de cáncer de mama

Abstract

A categorization system for the early identification of breast cancer has been developed using a convolutional neural network algorithm. The database includes 277,524 images of breast tissue, classified as healthy and cancerous, using preprocessing that ensures quality and stability. AlexNet was used with adjustments to hyperparameters such as batch size and learning rate to optimize performance. Data normalization and augmentation methods were applied to prevent overfitting. Evaluation metrics, such as detection capability and diagnostic accuracy, showed an accuracy of over 83% in the validation group. These results support the efficacy of AlexNet in the self-categorization of histopathological images, which will facilitate faster and more accurate diagnoses in the medical field.