Combination of contrast enhanced fuzzy c-means (CEFCM) clustering and pixel based voxel mapping technique (PBVMT) for three dimensional brain tumour detection

Abstract

Brain tumour is a lump of tissue produced by uncontrolled growth of cells in the human brain. Automated and accurate detection of brain tumour is important for robotics based surgery operation. The proposed method enhances the accuracy, sensitivity and specificity of automated brain tumour detection by adopting an accurate two phase method of detection. The method segments the Axial plane slices of Magnetic Resonance image in its first phase. In the second phase, binary decision values representing the presence or absence of tumor cells on each pixel locations are projected into 3D space to obtain the 3D tumor. Contrast Enhanced fuzzy c-means (CEFCM) clustering method is used to segment the 2D tumor regions from MR image slices due to its high accuracy. The decision values obtained from the segmented image for each pixel locations are mapped into 3D space using Pixel based voxel mapping technique (PBVMT). Average accuracy (Dice overlap coefficient) and average sensitivity of detection are measured with respect to the given ground truth of the image of BRATS 2013 dataset. The overall accuracy, sensitivity and specificity of detection are found to be 0.948, 92.14% and 96.97% respectively.