Brain Magnetic Resonance Image Tumor Detection and Segmentation Using Edgeless Active Contour

Abstract

Tumor detection and segmentation from a brain Magnetic Resonance Imaging (MRI) slice is a task of identifying a case of tumor, and masking out the tumor tissue region. It becomes a challenging task due to the complex distribution of the gray-scale intensities in the brain MRI. Additionally, skull stripping has been an essential pre-processing step, which removes the non-brain regions from the image. The existing skull stripping methods suffer from low accuracy and high complexity problems due to the variation in skull regions, depending upon the MRI slices obtained on different planes. These problems are solved in our work by selecting a region of interest using an Active Contour(AC), and using the symmetry characteristics of the human brain anatomy. The proposed method uses AC at multiple steps, and it is fully automatic in both tumor detection and segmentation. The algorithm is validated on a dataset of 3064 T1-weighted contrast-enhanced MRI images with three kinds of brain tumors. The method records detection efficiencies of 81%, 56% and 71% in meningioma, glioma and pituitary tumor classes respectively. Further, it exhibits a higher tumor segmentation performance as compared to a recent AC-based brain tumor segmentation method in each of the tumor classes separately.