Parkinson’s Disease Detection using Spiral Drawings and CNN



Original Source Here

The images are also reshaped to a common size (128, 128, 1). Images are further normalised before fitting the dataset to the model.

CNN Model Architecture

The implementation uses a CNN model architecture with the following characteristics —

  • The model contains four Convolutional Layers with 128, 64, 32, and 32 filters, respectively.
  • The convolutional layers contain filters with varying filter sizes.
  • A MaxPool2D layer follows each convolutional layer.
  • Two Fully Connected layers follow the convolutional block.

Defining the model using Keras —

The model summary is as follows —

Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv1 (Conv2D) (None, 128, 128, 128) 3328
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 40, 40, 128) 0
_________________________________________________________________
conv2 (Conv2D) (None, 40, 40, 64) 204864
_________________________________________________________________
max_pooling2d_5 (MaxPooling2 (None, 12, 12, 64) 0
_________________________________________________________________
conv3 (Conv2D) (None, 12, 12, 32) 18464
_________________________________________________________________
max_pooling2d_6 (MaxPooling2 (None, 4, 4, 32) 0
_________________________________________________________________
conv4 (Conv2D) (None, 4, 4, 32) 9248
_________________________________________________________________
max_pooling2d_7 (MaxPooling2 (None, 1, 1, 32) 0
_________________________________________________________________
flatten_1 (Flatten) (None, 32) 0
_________________________________________________________________
dropout_2 (Dropout) (None, 32) 0
_________________________________________________________________
fc1 (Dense) (None, 64) 2112
_________________________________________________________________
dropout_3 (Dropout) (None, 64) 0
_________________________________________________________________
fc3 (Dense) (None, 2) 130
=================================================================
Total params: 238,146
Trainable params: 238,146
Non-trainable params: 0
_________________________________________________________________

Training the Model

The model is trained with a learning rate of 3.15e-5 using Adam optimiser. The epochs and batch size are set to 70 and 128, respectively.

Performance of the Model

The Loss and Accuracy Plots, Classification Report and Confusion Matrix, are used as performance metrics for the model.

The Loss and Accuracy Plots are as follows —

AI/ML

Trending AI/ML Article Identified & Digested via Granola by Ramsey Elbasheer; a Machine-Driven RSS Bot

%d bloggers like this: