Alzheimer Stage Classifier

This is my first attempt creating Convolutional Neural Networks.I created a CNN to predict if a patient has Alzheimer's Disease and to classify the current Alzheimer stage based on patient's brain MRI scan The CNN has approximately 95% accuracy

Stages for classification

The neural network classifies a patient's brain MRI scan into the following categories

• Non Demented
• Very Mild Demented
• Mild Demented
• Moderate Demented

Dataset

The dataset used can me found here. I have merged the train and test directories found in the dataset , and split them using sklearn.modelselection.train_test_split to achieve better results in the training process.

Before you start

Before you start playing with the model run in the repo directory the following command to install the required packages for the model to run

$ pip install -r requirments.txt

Model Architecture

Convolutional Neural Network Architecture:

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d (Conv2D)              (None, 118, 118, 64)      640       
_________________________________________________________________
max_pooling2d (MaxPooling2D) (None, 59, 59, 64)        0         
_________________________________________________________________
conv2d_1 (Conv2D)            (None, 57, 57, 64)        36928     
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 28, 28, 64)        0         
_________________________________________________________________
flatten (Flatten)            (None, 50176)             0         
_________________________________________________________________
dense (Dense)                (None, 128)               6422656   
_________________________________________________________________
dropout (Dropout)            (None, 128)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 4)                 516       
=================================================================
Total params: 6,460,740
Trainable params: 6,460,740
Non-trainable params: 0
_________________________________________________________________

Training the model

To train the model all you have to do is to run :

$ python train.py

Make sure the data folder which contains the training data has not been altered in anyway

The model will be saved in the model directory with name "model.h5" overwriting the current pre-trained model.

Training Statistics

Model Accuracy

Model Loss

Using the model for making predictions

To use the model for making predictions first put brain MRI scans in the test directory

After, run :

$ python predict.py 

The script will load all the photos located in the test folder and will try to predict the Alzheimer stage based on the MRI scan

Updates and Feedback

I am looking forward to get your feedback on any issues that may occur. A new update is coming soon to improve the model's accuracy

License

All rights reserved.