Magnitude Normalized and OTSU Intensity based Brain Tumor Detection Using Magnetic Resonance Images

Abstract

Brain tumor is a deadly disease and its categorization is challenging due to the diversified characteristics of tumor cells. Currently, computer-aided techniques for brain tumor identification in the early stages are proposed with magnetic resonance imaging (MRI). In recent pre-trained models, features are obtained from bottom layers that vary from natural images to medical images. Moreover, attributable to disequilibrium and complication of noise and intensity of these lesions of brain tumors in MRI is yet considered to be the laborious strategy. To resolve these issues, an intensity-based segmentation method called, Magnitude Normalized Filtering and Otsu Intensity-based segmentation (MNF-OIS) is utilized. First, with the Brain MRI images provided as input, normalized pre-processed images are obtained by utilizing the Magnitude Normalized Saddle Median Filtering model. Next, Otsu Intensity-based segmentation is applied to the pre-processed image to obtain Convergence optimized segmented image. With the resultant segmented image testing on brain MRI images are made for brain tumor identification. MNF-OIS method is evaluated in terms of tumor detection time, convergence time, and accuracy. The experimental outcomes verify that the MNF-OIS method improved brain tumor detection accuracy by 7%, reduced the convergence time by 22%, and minimized tumor detection time by 32% as compared to existing Deep learning convolutional neural networks and DeepSeg, respectively.