A Hybrid Deep Learning Framework Integrating CBAM Attention and MobileNetV3 for Accurate Brain Tumor Detection

Abstract

Brain tumors are a life threatening health problem with high mortality rate and thus require proper detection procedures. This study will put forward a hybrid deep learning model to classify brain tumors using MRI scans. The algorithm combines MobileNetV3 to extract features effectively with a Convolutional Block Attention Module (CBAM) that adds channel-wise and spatial attention to emphasize diagnostically relevant areas. A Multilayer Perceptron (MLP) classifier, which consists of two hidden layers, is used to make a final classification of four classes: glioma, meningioma, pituitary tumor, and healthy tissue. The accuracy with a dataset of 7,000 MRI scans shows 98.93 percent with precision of 98.96 percent and recall of 98.93 percent.