A Robust Transfer Learning Approach with MobileNetV3-XAI for Multi-Class Brain Tumor Classification in MRI Images

Early tumor identification depends significantly on successful brain tumor classification to make effective treatment decisions. The research offers a deep learning algorithm which automates brain tumor sorting into four separate categories including, glioma, meningioma, pituitary tumor and no tumor. The MobileNetV3 system functions as an optimized feature extractor which retrieves complex MRI image representations at a high level. The network receives extracted features which are processed by a custom neural network head for classification purposes. The method performs evaluations using a publicly accessible brain tumor MRI dataset that includes 7153 MRI images. The suggested approach exhibits average performance value of 97.80% precision, recall, F1-score and accuracy. The classification results demonstrate the usefulness of the described approach for providing speedy medical decisions about brain tumor diagnoses to clinical staff.