Analysis of Fermi Filtering and Curvature-Intensified Random Walker-Based Feature Extraction for Brain Tumor Detection from Segmented MR Images

Abstract

Brain tumors represent a significant cause of mortality worldwide. Based on the origin and behavior of the cells, the World Health Organization (WHO) has identified 120 different types of brain tumors. Magnetic Resonance Imaging (MRI) is an essential tool for the early detection of brain cancers. Numerous techniques have been developed to extract tumor characteristics, but most of them have certain limits in terms of obtaining better PSNR and brain tumor detection accuracy in the shortest amount of time. This study employs Fermi Frequency Filtering and Curvature-intensified Random Walker Segmentation (FFF-CRWS) to improve feature extraction, achieve higher PSNR, and accelerate diagnosis from MRIs. To eliminate noisy artifacts from brain MRIs, preprocessing employs Fermi Frequency Filtering and Data Density Functional Theory (DDFT). Using Curvature-intensified Random Walker Segmentation, the preprocessed image is divided into image segments. To effectively distinguish between tumor and healthy brain tissue, pertinent features, such as texture (e.g., intensity variations, homogeneity), shape (e.g., area, perimeter, and eccentricity), spatial (e.g., location and boundary characteristics), and intensity, are extracted for tumors from segmented MRIs by examining the intensity variations of pixels within the tumor boundary.