NeuroFusion 🧠

Overview

NeuroFusion is a full-stack AI-based diagnostic support system designed to detect Alzheimer’s Disease from structural MRI scans. It integrates Vision Transformers (ViT) with a modern web interface to provide real-time predictions and explainable AI outputs.

The system is built to address the critical challenge of early Alzheimer’s detection, where brain degeneration begins 15–20 years before clinical symptoms appear.

Problem Statement

Traditional diagnostic methods and CNN-based AI models struggle with:

• Early-stage Alzheimer’s detection
• Poor generalization across medical institutions
• Lack of explainability in predictions
• Limited ability to capture global brain structure patterns

Key Features

• MRI scan upload and preprocessing pipeline
• AI-based prediction (CN / MCI / AD classification)
• Vision Transformer-based inference engine
• Explainable AI heatmaps (attention visualization)
• Real-time web dashboard
• User authentication system

Tech Stack

System Architecture

NeuroFusion follows a modular architecture:

• Frontend: Next.js web application
• Backend: Node.js + Express API layer
• AI Engine: Python-based Vision Transformer model
• Database: MongoDB for user and prediction storage

Research Foundation

The system is built on extensive literature review of Vision Transformers in medical imaging, achieving reported diagnostic performance of 92.5% sensitivity and 95.7% specificity in AD detection. It leverages insights from ADNI, OASIS, and UK Biobank datasets.

Impact

NeuroFusion aims to support early clinical decision-making in Alzheimer’s diagnosis, enabling earlier intervention and improved patient outcomes through AI-assisted analysis.