An RCNN and Gaussian Mixture Approach to MRI Tumor Classification and Detection

Brain tumour identification and analysis is an essential component of any indication system, as proven by numerous studies and the steady advancement of relevant methodologies. Since this study relies on accurate tumour detection, a reliable automated method is required. For more accurate brain tumour classification, several segmentation algorithms have been created. Brain image segmentation is notoriously tough, according to specialists in medical image processing. The authors of this paper offer a new approach to automated detection and classification. The proposed approach is divided into multiple steps, such as MRI image preprocessing, image segmentation, feature extraction, and feature classification. As part of the preprocessing phase, an adaptive filter is used to decrease the noise in the MRI image. For feature extraction, we use the grey level co-occurrence matrix (GLCM), and for picture segmentation, we use the improved Means clustering (IMC) method. Following feature extraction from MRI scans, we used a deep learning system to classify scans as either benign lesions, gliomas, meningiomas, or pituitary tumours. The classification job was completed with the help of recurrent convolutional neural networks (RCNN). The suggested method outperforms the state-of-the-art in terms of classification for every given input dataset that contains brain images. In this study, MRI images from 464 testing sets and 3264 training sets were used from Kaggle's data sets. According to the results, the new approach is much better than the old ones. Three of the most widely used classification algorithms at the moment—BP, VG-Net, and RCNN—are compared to the proposed RCNN method. In MRI scans of brain tumours, the proposed classifier correctly identified 97.81 percent of the tissues.