CNN Model for Alzheimer's MRI Dataset

This project implements a Convolutional Neural Network (CNN) for classifying MRI scans into various stages of Alzheimer's Disease using the Best Alzheimer's MRI Dataset.

Dataset

The dataset comprises high-quality MRI scans labeled for Alzheimer's Disease progression. It can be downloaded from Kaggle.

Problem Statement

The task is a classification problem to predict the stage of Alzheimer's Disease.

Classes

The dataset includes the following classes:

• Mild Impairment
• Moderate Impairment
• No Impairment
• Very Mild Impairment

Model Architecture

The CNN model comprises two convolutional blocks followed by a fully connected classifier; influenced by TinyVGG16 architecture.

Performance Metrics

The model was evaluated on both training and test datasets. Below are the key metrics:

Training Metrics

• Loss: 0.0039
• Accuracy: 99.90%
• Precision: 0.9990
• Recall: 0.9990
• F1-Score: 0.9990

Testing Metrics

• Loss: 0.1574
• Accuracy: 95.47%
• Precision: 0.9563
• Recall: 0.9547
• F1-Score: 0.9548

Train-Test Loss Over Epochs:

Train Data Confusion Matrix:

Test Data Confusion Matrix:

How to get the code

git clone https://github.com/Recker-Dev/alzheimer-cnn-tinyVGG16.git
cd alzheimer-cnn

How to Setup Environment

Follow these steps to set up and run the project:

1. Create a Conda environment
   Run the following command to create a new environment named alz_torch_cnn with Python 3.10:

   conda create --name alz_torch_cnn python=3.10 -y

2. Activate the environment

   conda activate alz_torch_cnn

3. Install dependencies

   pip install -r requirements.txt

4. Deactivate the environment

   conda deactivate

How to Use the Model

You can interact with the model in two ways:

1. Using Jupyter Notebooks
   There are notebooks provided under the notebooks section. These notebooks allow you to experiment with the model, visualize predictions, and understand its workings in a step-by-step manner.

2. Using the Streamlit Application
   For users without much coding background, an application (app.py) is provided for easy inference. Follow these steps to use the app:

   • Activate the environment:

     conda activate alz_torch_cnn

   • Run the Streamlit app:

     streamlit run app.py

3. Testing Images
   Testing images are provided in the Sample Testing Alzheimer Dataset folder. Use these images to test the model's performance through the Streamlit app or the provided notebooks.

Folder Structure:

root
├── app.py        # Main Streamlit application for user interaction
├── models
│   └── alz_CNN.pt  # Trained PyTorch model for Alzheimer's classification
├── model_arch.py  # Defines the architecture of the Alzheimer's classification model
├── notebooks      # Jupyter notebooks for development and analysis
│   ├── Alzehmier_CNN.ipynb  # Notebook for training the Alzheimer's classification model
│   └── predict_using_alzehmierCNN.ipynb  # Notebook for demonstrating model prediction
└── Sample Testing Alzehimer Dataset  # Folder containing sample images for testing
    └── test            # Subfolder containing images for inference/testing
        ├── Mild Impairment
        ├── Moderate Impairment
        ├── No Impairment
        └── Very Mild Impairment
└── requirements.txt  # File listing required Python libraries