Brain Tumor Segmentation based on Rough Set Theory for MR Images with CA Approach

Abstract

Medical imaging as a science has advanced significantly over the past few decades Medical imaging has grown in popularity as a result of increasingly advanced equipment and non-invasive procedures, which has resulted in more accurate diagnosis. However, the fundamental principle of effective diagnosis is the acquisition of noise-free images, which remains elusive. Parallel research is being conducted on eliminating artifacts/noise caused by hardware, software, and physical problems. This is an effort made in the present thesis. For various noise models, techniques for estimating the original data from its noisy form are provided. Magnetic Resonance Imaging (MRI) of brain tissues is the type of medical image considered. newlineThe newly defined membership function creates an intuitiveistic fuzzy image that accounts for the inherent inhomogeneity of intensity, inter-region heterogeneity and thermal noise in MR images. For the segmentation of MR images, a novel approach predicated on intuitionistic fuzzy rough set theory is presented. Rather than using a histogram to determine segmentation thresholds, this algorithm uses an intuitionistic fuzzy roughness index to ascertain the optimal segmentation thresholds for MR images. The intuitionistic fuzzy roughness index is computed using a lower and an upper approximation, histogram and histon, respectively. The intuitiveistic fuzzy roughness index identifies significant peak and valley points in the image that correspond to homogeneous regions. Multilevel thresholding is used to segment the MR images into three phases based on these significant valley points: medulla, cortex, and pelvic regions. However, only the medulla and cortex are clinically significant of these three segments. newlineA variety of validation methods, including Jaccard similarity (JS), Dice coefficient (DC), and confusion table, are used to assess the segmentation results of the proposed methods for segmenting brain MR images.