In this study, we provide a Convolutional Neural Network-based model for fully autonomous segmentation and classification of brain tumors. Our concept differs from other efforts in that the input images are processed along various processing pathways at three different spatial scales. The natural functioning of the human visual system served as inspiration for this technique. The suggested neural model does not require pretreatment of input pictures in order to remove portions of the skull or vertebral column beforehand. It can interpret MRI scans containing Meningioma, glioma, and pituitary tumors in sagittal, coronal, and axial perspectives. Worldwide, there is a high fatality rate from brain tumors. Medical professionals must visually analyze the images and mark out the tumor areas, a process that takes time and is prone to error. Automated methods for early brain tumor identification have recently been developed.) However, due to their high false-positive outcomes and low accuracy, these approaches have issues. An efficient tumor identification and classification approach is required to extract robust characteristics and perform accurate disease classification.) This study proposes a novel deep feature-based multiclass classification method for brain tumors.) Min-max normalization is used as a preprocessing step for MR images before considerable data augmentation is undertaken to address the data problem. A single feature vector created by combining deep CNN features from transfer learned architectures such as AlexNet, GoogleNet, and ResNet18 is then loaded into SVM and K-nearest neighbor (KNN) to make a prediction. The suggested method's classification effectiveness is enhanced by the unique feature vector, which contains more data than the independent vectors. The recommended method also performed better than the existing methods and produced good accuracy; as a result, it may be used in a clinical context to categorise brain tumors using MRIs.

Brain Tumor, Classification, Segmentation, Convolutional Neural Network, MRI, Artificial Intelligence