An optimized dual attention-based network for brain tumor classification

Abstract

Brain tumors are one of the leading causes of death worldwide. Different types of brain tumors are known, so the choice of treatment depends directly on the type of tumor. The classification of brain tumors is very important as a complex and challenging problem in the field of image processing. Today, deep learning methods are used to classify brain tumors. In addition to being able to detect and automatically classify all types of brain tumors, these methods significantly reduce the diagnosis time and increase accuracy. In this paper, a deep learning-based model is proposed to classify brain tumors into three classes: glioma, meningioma, and pituitary tumor. In the first phase, the pre-trained network ResNet50 is used to extract features from MRI images. In the second phase, by proposing two attention mechanisms (depth-separable convolution-based channel attention mechanism and an innovative multi-head-attention mechanism), the most effective spatial and channel features are extracted and integrated. Finally, the classification phase is performed. Evaluations on the Figshare dataset showed an accuracy of 99.32%, which performs better than existing models. Therefore, the proposed model can accurately classify brain tumors and help neurologists and physicians make accurate diagnostic decisions.