Quantitative Analysis of Fuzzy Roughness Measure based Segmentation of Magnetic Resonance Brain Images

Abstract

Quantitative analysis of segmentation results in image processing using various quantitative indices enable to prove superiority of method proposed. Visualizing brain tumors and diagnosing brain diseases is an important application of magnetic resonance imaging (MRI). The brain disease study requires segmentation of brain tissue into nonoverlapping regions of cerebral brain tissues. The white matter (WM) volume estimation is an indicator for the diagnosis of Alzheimer’s diseases. Gray matter (GM) controls the sensory perception of human. Cerebrospinal fluid (CSF) provides the protection. Hence this segmentation is very important for neuro disease diagnosis purpose. But segmentation of these brain tissues faces problems due to intensity variation and noise. The fuzzy roughness-based thresholding method is proposed in this paper. The method detects the optimum thresholding level by exploiting the spatial relationship between the pixel intensities near the boundary of brain tissues. Method proposed in this paper partitions brain MR image into non-overlapping regions. Quantitative analysis of the segmentation result is important to prove the accuracy of the obtained results against ground truth. Hence, performance evaluation of the proposed method is quantified using Jaccard Similarity (JS), Dice coefficient (DC), and under, over and incorrect segmentation.