Kernelized fuzzy C means with recurrent convolutional network for brain tumor detection and classification from 3D MRI image

Detection of brain tumors with 3D MRI entails using three-dimensional magnetic resonance imaging data to identify and characterize abnormal brain growths. However, in current brain tumor detection techniques, frequently suffer from segmentation issues and spatial information loss during classification, because voxel-based representations fail to capture the spatial relationships in the complex tumor structure. To address these challenges, a novel "Kernelized Fuzzy C Means with Recurrent Convolutional Network," has been proposed. The process begins with pre-processing, which reduces noise and artifacts in 3D MRI images. Then Adaptive Watershed Local Kernelized Fuzzy C Means is employed for segmentation, which effectively addresses low contrast challenges. After segmentation, an novel approach known as Sobel Pyramid Transform with Recurrent Convolutional Neural Network is implemented to evolve feature extraction-based classification, where Sobel-based Steerable Pyramid Transform is utilized for feature extraction, which extracts spatial relations and context in tumor regions, as well as textural qualities and for classification 3D Randomized Weighted-based Decisive Recurrent Convolutional Network is used, which captures and use spatiotemporal information to accurately and robustly predict brain tumor regions of cancer. Finally, the proposed model is executed in MATLAB, where it reveals superior performance by achieving a higher accuracy of 99%, IoU of 98.71% with a reduced computation time of 2.7 s for precise detection and classification of brain tumor.