AI Pipeline for Epilepsy Detection

This project represents a comprehensive AI pipeline developed for epilepsy detection and presurgical decision-making. The system leverages arterial spin labeling MRI data combined with foundational models to provide accurate neurological disease prediction.

Key Features

• Advanced MRI Analysis: Utilizes arterial spin labeling techniques for improved brain perfusion analysis
• Foundational Models: Integration of large-scale pre-trained models for neurological disease prediction
• Clinical Decision Support: Provides actionable insights for presurgical planning
• Adaptive Learning: Implements metadata-guided supervised contrastive learning for improved performance

Technical Implementation

The pipeline combines multiple AI techniques including:

• Convolutional Neural Networks for MRI image analysis
• Transformer architectures for sequence modeling
• Contrastive learning for robust feature representation
• Multi-modal fusion for comprehensive analysis

Clinical Impact

This work contributes to improved patient outcomes by:

• Enhancing accuracy of epileptic focus detection
• Supporting presurgical planning decisions
• Reducing time to diagnosis
• Providing quantitative metrics for clinical assessment

Funding & Recognition

• Supported by Compute Canada RAC grant (55k$ & 45 RGU-years GPU)
• Published in IEEE Journal of Biomedical and Health Informatics
• Collaboration with Queen’s University Centre for Neuroscience Studies