Segmentation of brain tumors using a semi-automatic computational strategy

Abstract

In this work, a semi-automatic computational strategy is proposed for brain tumor segmentation. The filtering (erosion + gaussian filters), segmentation (level set technique) and quantification (BT volume) stages are applied to magnetic resonance imaging in order to generate the three-dimensional morphology of brain tumors. The Jaccard's Similarity Index is considered to contrast manual segmentation with semi-automatic segmentations of brain tumor. In this sense, the highest Jaccard's Similarity Index provides the best parameters of the techniques that constitute the semi-automatic computational strategy. Results are promising, showing an excellent correlation between these segmentations. The volume is used for the brain tumors characterization.