alzheimer-cbam-cnn

Alzheimer Disease MRI Classification (CBAM CNN)

This model uses a custom Convolutional Neural Network (CNN) integrated with CBAM (Convolutional Block Attention Module) attention mechanisms. It distinguishes between four classes: Mild Demented, Moderate Demented, Non Demented, and Very Mild Demented.

Model Architecture

• Backbone: 4-block Custom CNN
• Attention: Channel & Spatial Attention (CBAM) in each block
• Input: 224x224 RGB images

Performance (Test Set)

• Accuracy: 0.6375
• Weighted Average F1-Score: 0.6255

Classification Report:

                    precision    recall  f1-score   support

     Mild Demented       0.67      0.32      0.43       172
 Moderate Demented       0.00      0.00      0.00        15
      Non Demented       0.72      0.79      0.75       634
Very Mild Demented       0.52      0.57      0.54       459

          accuracy                           0.64      1280
         macro avg       0.48      0.42      0.43      1280
      weighted avg       0.63      0.64      0.63      1280