Localização de fraturas na coluna cervical em CT utilizando redes neurais convolucionais.

Abstract

Worldwide, every 5 seconds a spinal column vertebrae is fractured and with the natural population aging process this time interval tends to reduce. The main method for fracture analysis is the computed tomography (also known as CT) and the time required for its analysis is crucial to avoid complications to the patients. Therefore the usage of techniques such as convolutional neural networks may accelerate the healthcare professional's decision making process. This article aims at investigating a convolutional neural network approach to localize fractures in CT slices. The RSNA (Radiological Society of North America) 2022 fracture detection database, made available in the Kaggle platform, was explored with the purposes of training and validating a proposed neural network. It is expected that the investigation conducted in the work and the model produced may assist health professionals in the task of detection and localization of fractures in the cervical column.