Concept Explainable Fusion of Bayesian CNN, ResNet and Vision Transformer for Brain Disorder Staging

This project presents an explainable model that combines Bayesian Convolutional Neural Networks (CNN), ResNet, and Vision Transformers (ViT) for brain disorder staging. The goal is to classify brain disorder stages using brain imaging data, specifically from MRI scans, and to provide explainable results to enhance interpretability in medical diagnoses. The dataset used for this study is sourced from Kaggle, which contains labeled brain images categorized into four stages: Non-Demented, Very Mild Demented, Mild Demented, and Moderate Demented. Bayesian CNN is employed to handle uncertainty in the predictions, while ResNet and Vision Transformer models are leveraged for their deep learning capabilities in capturing complex patterns in brain images. The integration of these models results in a robust and accurate classifier for brain disorder stages. The system is built using Flask for the backend, with a front-end interface developed in HTML, CSS, and JavaScript. This approach aims to improve diagnostic accuracy and provide reliable, interpretable results for clinical practitioners.

Keywords:
Brain disorder staging, Bayesian CNN, ResNet, Vision Transformer, MRI scans, dementia classification, deep learning, medical imaging, machine learning, explainability, Flask, AI in healthcare.