Brain Tumor Detection and Classification Using Attention Based Residual Learning

In this paper, we present a novel approach for brain tumor detection and classification. This may prove essential for the evolving field of medicine. The work helps to not just identify the major type of tumor but also detect the localized region of tumor in the brain from the MRI images of the patient's brain. This can assist and perhaps expedite the analyses done by and the interventions made by doctors and medical professionals to make further decisions for the specific patient. For this reason, we propose a hybrid model comprising of a deep learning model leveraging the capabilities of ResNet-50, an attention module, and a neural network model of dense layers. This model is used to classify the brain tumor as either a glioma, meningioma, or pituitary type, or as a brain image with no tumors. Our model gives a promising solution for automated brain tumor detection and classification. This has the potential to help in early detection of malignant tumor types and offset or prevent further disabilities that would otherwise be caused by such tumors.