Brain tumor MRI images identification and classification based on the recurrent convolutional neural network

Abstract

Brain tumor detection and analysis are necessary for any indicative system and have testified that exhaustive research and procedural development over time. This work needs to implement an effective automated system to improve the accuracy of tumor detection. Various segmentation algorithms have been developed to achieve and enhance the accuracy of brain tumor classification. Brain image segmentation has been recognized as a complex and challenging area in medical image processing. This paper proposes a novel automated scheme for detection and classification. The proposed method is divided into various categories: MRI image preprocessing, image segmentation, feature extraction, and image classification. The image preprocessing step is performed using an adaptive filter to remove the noise of the MRI image. Image segmentation is performed using the improved K-means clustering (IKMC) algorithm, and the gray level co-occurrence matrix (GLCM) is used for feature extraction to extract features. After extracting features from MRI images, we used a deep learning model to classify the types of images such as gliomas, meningiomas, non-tumors, and pituitary tumors. The classification process was performed using recurrent convolutional neural networks (RCNN). The proposed method provides better results for classifying brain images from a given input dataset. The experiments were conducted on the Kaggle dataset with 394 testing sets and 2870 training set MRI images. The results illustrate that the proposed method achieves a higher performance than previous methods. Finally, the proposed RCNN method is compared with the current classification methods of BP, U-Net, and RCNN. The proposed classifier obtained 95.17% accuracy in classifying brain tumor tissues from MRI images.