Hybrid optimization based deep learning method for glioma brain tumor detection using mri

Abstract

Brain tumor segmentation and classification is a complex in the medical imaging system, as treatment planning and diagnosing the tumor is difficult. Magnetic Resonance Imaging (MRI) is specially used to assess the gliomas, as it offers complementary information for acquiring the MRI sequences. To label the brain tumor accurately with the associated edema using MRI is significantly a time-consuming process, and observed considerable variation between the labelers. Different classification and segmentation methods are introduced in the past few decades to accurately classify the abnormal tissues from brain. The most reported challenge by most of the tumor classification and segmentation mechanism is regarding the accurate segmentation and classification of tumor region, which is addressed in this research through exploiting the effective tumor segmentation method. newlineAccordingly, in this research three different tumor classification methods are presented. In the first contribution, deep joint segmentation is introduced to perform edema and core tumor segmentation. Here, brain images are partitioned to grids, and the pixels are formed by computing the mean and threshold factors. The region fusion is performed based on the bi-constraints and region similarity, and the optimal segments are determined based on the distance between the segmented points and the deep points. newlineThe second contribution is based on Fractional Jaya Optimizer-based Deep Convolutional Neural Network (FJO-DCNN) for achieving tumor classification using the segmented features of MRI. newline