Transfer deep learning and explainable AI framework for brain tumor and Alzheimer's detection across multiple datasets

Abstract

IntroductionThe pressing need for accurate diagnostic tools in the medical field, particularly for diseases such as brain tumors and Alzheimer's, poses significant challenges to timely and effective treatment.MethodsThis study presents a novel approach to MRI image classification by integrating transfer learning with Explainable AI (XAI) techniques. The proposed method utilizes a hybrid CNN-VGG16 model, which leverages pre-trained features from the VGG16 architecture to enhance classification performance across three distinct MRI datasets: brain tumor classification, Alzheimer's disease detection, and a third dataset of brain tumors. A comprehensive preprocessing pipeline ensures optimal input quality and variability, including image normalization, resizing, and data augmentation.ResultsThe model achieves accuracy rates of 94% on the brain tumor dataset, 81% on the augmented Alzheimer dataset, and 93% on the third dataset, underscoring its capability to differentiate various neurological conditions. Furthermore, the integration of SHapley Additive exPlanations (SHAP) provides a transparent view of the model's decision-making process, allowing clinicians to understand which regions of the MRI scans contribute to the classification outcomes.DiscussionThis research demonstrates the potential of combining advanced deep learning techniques with explainability to improve diagnostic accuracy and trust in AI applications within healthcare.