Utilização de transfer learning em redes neurais convolucionais: Identificação de Alzheimer em imagens de ressonância magnética

Abstract

The Alzheimer’s Disease is the main cause of dementia in the world and due to it not having a cure, the treatment must be started as soon as possible to minimize the symptoms The objective of this paper is to apply transfer learning techniques to a convolutional neural network (CNN), aiming to identify dementia patterns in Magnetic Resonance Imaging and assist in Alzheimer’s Disease diagnotics. Applying the network can help the clinical decision making, by reducing false negatives of anomalies that might not be noticed at first. The model was developed using VGG19 as base and Python libraries such as Tensorflow and Augmentor. The model reached an accuracy of 98.44% on the validation.