NeuroXAI

NeuroXAI provides explainability for deep neural networks applied to brain MRI analysis by generating visualization maps from multiple explainable AI methods to interpret classification and segmentation models for brain tumor assessment.

Key Features:

• Explainable AI framework: Integrates seven state-of-the-art explanation methods to generate visualization maps that elucidate how deep learning models make decisions.
• Image classification and segmentation: Applied to MRI-based image classification and segmentation tasks focused on brain tumor analysis.
• Visualization techniques: Produces visual attention maps from multiple XAI methods and visualizes information flow within internal layers of segmentation convolutional neural networks (CNNs).
• Comparative analysis of explanations: Enables side-by-side comparison of multiple XAI methods to aid interpretation of model behavior.
• Open architecture and scalability: Designed to incorporate new XAI methods and scale the explanation framework as methods evolve.

Scientific Applications:

• Brain tumor detection and diagnosis: Supports interpretation of model predictions for MRI-based detection and diagnosis of brain tumors.
• Clinical decision support in neuro-oncology: Aids clinicians and radiologists in understanding model outputs to inform diagnostic interpretation and treatment planning.

Methodology:

Integrates seven explanation methods to generate visualization maps, produces visual attention maps from multiple XAI methods for comparative analysis, and visualizes information flow within internal layers of segmentation CNNs.