To develop a new NeuroXAI framework for obtaining 2D and 3D explainable sensitivity maps to assist clinicians in understanding and trusting deep learning algorithms in clinical procedures.
Key Findings:
Deep learning techniques, particularly convolutional neural networks, have shown promise in brain tumor segmentation and classification.
Explainable AI (XAI) methods are essential for understanding deep learning predictions in sensitive medical applications.
Current XAI methods have limitations, primarily focusing on 2D MRI slices rather than 3D applications.
Interpretation:
The need for explainability in deep learning models is critical in medical imaging to ensure trust and compliance with regulations like GDPR.
Limitations:
Existing studies primarily focus on 2D MRI slices, lacking comprehensive 3D analysis.
Explainability methods may trade off model complexity for performance.
Conclusion:
The NeuroXAI framework aims to bridge the gap in explainability for deep learning models in brain imaging, enhancing clinician trust and understanding.
