Sistema para treinamento e classificação de imagens de radiografias

Abstract

Chest X-ray is the most common type of imaging examination in the world. It is widely used for diagnosing thoracic diseases and is one of the most valuable tools when interpreted carefully. In this context, to assist doctors in diagnosing thoracic diseases more easily, this study proposes a machine learning model to perform such diagnoses using cloud-based X-ray images. Accuracy, loss, and area under the curve (AUC) metrics are used to evaluate the model’s performance. The proposed model demonstrates a high success rate, achieving 98% accuracy for some diseases. Additionally, this study implemented a system provisioned in an AWS architecture, setting it apart from other works related to X-ray classification.