A Hybrid Deep Learning Approach For Brain Tumor Detection Using XAI with GradCAM

The abnormality of cells in the brain causes brain tumors. If the brain is left untreated, later on it exposes as cancer. Globally, it has become one of the highest mortality among adults. Effective and timely identification of brain tumors with a proper treatment plan can improve a patient's quality of life and extend life expectancy. In MRI (Magnetic Resonance Image), segmentation of brain tissues is crucial for detecting the presence of outlines related to any brain tumor. Deep Learning (DL) and Explainable AI (XAI) have recently gained significant attention as promising methods for detecting brain tumors. This research carried out deep learning based two-hybrid models such as VGG19+LSTM and VGG16+INCEPTIONV3 for detecting brain tumors, and for reliability and explainability, we are using XAI technique called GradCAM on Brain Tumor MRI dataset from Kaggle. The dataset contains more than 7000 images, including four classes such as glioma, meningioma, pituitary and notumor. The evaluation parameters for model performance are Accuracy, Precision, Recall, and F1-score. The outcomes of this study indicate that VGG19+LSTM outperforms the other model, with an accuracy of 98%. And for VGG16+INCEPTIONV3, we got an accuracy of 96%. Then, we applied GradCAM to the best classifier, VGG19+LSTM.