Transfer deep learning and explainable AI framework for brain tumor and Alzheimer's detection across multiple...

Area of Science:

• Medical Imaging Analysis
• Artificial Intelligence in Healthcare
• Neurological Disease Diagnostics

Background:

• Accurate diagnosis of neurological conditions like brain tumors and Alzheimer's disease is critical for effective treatment.
• Current diagnostic methods face challenges in speed and accuracy, necessitating advanced tools.

Purpose of the Study:

• To develop and evaluate a novel approach for MRI image classification using transfer learning and Explainable AI (XAI).
• To enhance diagnostic accuracy and provide transparency in AI-driven medical image analysis.

Main Methods:

• A hybrid CNN-VGG16 model was employed for MRI image classification.
• Transfer learning utilizing VGG16 pre-trained features was integrated.
• Three MRI datasets (brain tumors, Alzheimer's disease, brain tumors) were used, with preprocessing including normalization, resizing, and data augmentation.
• SHapley Additive exPlanations (SHAP) were incorporated for model interpretability.

Main Results:

• The model achieved high classification accuracies: 94% for brain tumors, 81% for Alzheimer's disease, and 93% for the second brain tumor dataset.
• SHAP analysis provided insights into the decision-making process, highlighting critical scan regions for classification.

Conclusions:

• The combination of deep learning and XAI significantly improves diagnostic accuracy for neurological conditions.
• This approach enhances clinician trust in AI applications for medical diagnostics by offering transparent decision-making processes.