Segmentação de lesões cerebrais utilizando deep learning em imagens de ressonância magnética

Abstract

This study investigated the application of Deep Learning techniques in the segmentation of brain lesions in MRI images, with a focus on evaluating how different training approaches influence model effectiveness. The experiments were conducted with three variants of a dataset, each one subjected to different pre-processing and data augmentation techniques. The research concluded that accuracy, performance and the nature of the data are interconnected, affecting the final result in complex ways. The findings emphasize the importance of training planning and the need to balance multiple techniques to optimize results. While the potential of deep learning in segmenting brain lesions is clear, excellence in this area requires a methodical and adaptive approach, optimally combining available techniques and configurations.