Automatic segmentation of a meningioma using a computational technique in magnetic resonance imaging

Resumen

Through this work we propose a computational techniquefor the segmentation of a brain tumor, identified as meningioma(MGT), which is present in magnetic resonance images(MRI). This technique consists of 3 stages developed inthe three-dimensional domain: pre-processing, segmentationand post-processing. The percent relative error (PrE) is consideredto compare the segmentations of the MGT, generatedby a neuro-oncologist manually, with the dilated segmentationsof the MGT, obtained automatically. The combination ofparameters linked to the lowest PrE, provides the optimal parametersof each computational algorithm that makes up theproposed computational technique. Results allow reporting aPrE of 1.44%, showing an excellent correlation between themanual segmentations and those produced by the computationaltechnique developed.