An Improved Convolutional Neural Network Framework for Brain MRI Image Classification

The number of deaths due to brain tumors is increasing day by day. Processes such as early diagnosis and treatment planning require time and expertise. As in every field, the development of artificial intelligence-based systems in the field of health contributes to the diagnosis of diseases. In this study, a new Convolutional Neural Network (CNN) model is proposed for the detection of brain tumors in MRI images that will provide better results than the studies in the literature. The proposed model consists of a modified Inception block, an efficient channel attention block (ECA) and an Atrous spatial pyramid pooling module (ASPP). Feature extraction is performed with the modified Inception block and these extracted features are enhanced with the ECA block. Features with informative context information are obtained with the ASPP module. Finally, classification is performed using convolution layers and Softmax layer. When the model is trained and tested with Figshare dataset, 99.80% accuracy is achieved.