Algoritmo semi-automático para análise e segmentação de tumores cerebrais utilizando técnicas básicas de processamento digital de imagens

Abstract

Brain tumors, particularly gliomas originating from glial cells, represent a significant challenge for medicine due to their complexity and high morbidity and mortality rates. The diagnosis of this disease primarily relies on magnetic resonance imaging (MRI), a technique widely used in the medical field as it eliminates the need for ionizing radiation and provides detailed, high-quality images of any region of the body. In this context, a semi-automatic algorithm was developed based on basic digital image processing techniques for the segmentation and feature extraction of brain tumors in MRI glioma-type images. The public database used was Kaggle, consisting of 3,264 images featuring three tumor types: gliomas, meningiomas, and pituitary tumors. Using the strategy developed and applied to glioma-type tumors, it was possible to achieve an accuracy rate of 77.47% in identifying relevant tumor regions. Despite the limitations associated with the need for manual adjustments, the results demonstrate the potential of the approach to provide support for more accurate diagnoses and therapeutic planning.