Densely Connected CNN-Based XAuNet1.0 for Brain Tumor Classification via MRI Images in the IoT Era

Brain tumor classification stands as a pivotal challenge within the realm of Computer-Aided Diagnosis. This research delves into the binary classification of brain images derived from varied angles in Magnetic Resonance Imaging scans of the human brain. The proposed classification model is built on deep transfer learning paradigm, implementing pre-trained Densely Connected Convolutional Neural Networks and “Image Net” weights to autonomously extract features from the inputs of MRI brain images based on DenseNets. Our model was trained on a large dataset of two classes “Yes” and “No” taking into consideration the importance of hyper parameters tuning role and the usage of regularization techniques effectively. Therefore, our results suggest that our proposed model XAuNet1.0 performs exceptionally well in accurately identifying instances of both classes, making it a robust solution for real-world tests and live usage through deploying the model in an accessible platform or IoT health care systems.