XAI and Computer Vision: Interpretable Deep Learning Approach with Explainable Artificial Intelligence for Brain Tumor Classification

Explainable Artificial Intelligence (XAI) plays a crucial role in the field of healthcare by enhancing transparency and decision-making processes in computer vision applications. In our research, we utilized two Convolutional Neural Network (CNN) models, SeparableConvNet and ConvNet, for the classification of brain tumors using computer vision techniques. These models were evaluated using a dataset consisting of 7023 human brain MRI scans collected from figshare, SARTAJ and Br35H datasets on Kaggle. The images were categorized into four groups: Gliomas, meningiomas, no tumor and pituitary. Both models underwent preprocessing steps including data normalization and augmentation to improve reliability and generalizability. The SeparableConvNet achieved a test accuracy of 96.64%, outperforming the ConvNet's test accuracy of 95.19%. To enhance interpretability and gain insights into the models' decision-making processes, “Local Interpretable Model-agnostic Explanations (LIME)” were employed. Our results highlight the effectiveness of CNN models in classifying brain tumors using computer vision and underscore the importance of XAI techniques in understanding these models' predictions.