Low grade glioma segmentation using an automatic computational technique in magnetic resonance imaging

Abstract

Through this work we propose a computational technique forthe segmentation of a brain tumor, identified as low gradeglioma (LGG), specifically grade II astrocytoma, which ispresent in magnetic resonance images (MRI). This techniqueconsists of 3 stages developed in the three-dimensionaldomain. They are: pre-processing, segmentation and postprocessing.The percent relative error (PrE) is considered tocompare the segmentations of the LGG, generated by a neuro-oncologist manually, with the dilated segmentations of theLGG, obtained automatically. The combination of parameterslinked to the lowest PrE, allow establishing the optimal parametersof each computational algorithm that makes up theproposed computational technique. The results allow reportinga PrE of 1.43%, which indicates an excellent correlationbetween the manual segmentations and those produced bythe computational technique developed.