ADHD-EEG-signal-classification

Our project is Binary EEG signal classification using 1D CNN deep learning model
We are classifying our signal into control and ADHD\

Table of content

- Dataset

- Data conversion

- Data preprocessing

- Data preparation and 1D CNN model

- Confusion matrix

- Metrics

- ROC curve

First: Dataset

• the dataset was aquired from IEEEDataPort EEG data for ADHD / control children \
• Participants were 61 children with ADHD and 60 healthy controls (boys and girls, ages 7-12).
• The ADHD children were diagnosed by an experienced psychiatrist to DSM-IV criteria, and have taken Ritalin for up to 6 months.
• None of the children in the control group had a history of psychiatric disorders, epilepsy, or any report of high-risk behaviors.
• EEG recording was performed based on 10-20 standard by 19 channels (Fz, Cz, Pz, C3, T3, C4, T4, Fp1, Fp2, F3, F4, F7, F8, P3, P4, T5, T6, O1, O2) at 128 Hz sampling frequency.
• Since one of the deficits in ADHD children is visual attention, the EEG recording protocol was based on visual attention tasks.

Second: Data conversion

Converting files

• The data files was in .mat format, we converted it to .fif to help us deal with it and saved it

Third: Data preprocessing

• we created new file pathes one for ADHD and one for Control to save the filtered data in
• We used MNE library for reading, diplaying and preprocessing files
• We applied high-pass filter at frequency = 0.5 Hz
• Then, We applied notch filter at frequency = 50 Hz to remove power line frequency
• After that, We applied zero padding to all the files to avoid the loss of the data at the begining and end of record
• Then, We applied moving average filter with window = 5, the aim of it is to smooth out the signal a bit
• Then, We applied common average reference filter to remove artifacts related to noise present during the record
• At the end, We saved the filtered data in the files

Fourth: Data preparation and 1D CNN model

Data preparation

• we read the paths of the filtered data first
• we took 50 ADHD files and 60 control files to make the data balanced in both classes as the time of ADHD files were longer so it'll produce more epoches
• we created a function called read_data:
  • first it reads the files from the new directory
  • then, it divides single record into multiple equal epochs, epoch duration = 10 seconds with 75% overlap
  • then, it converts each epoch to numpy array using epoch.get_data()
• after applying this function to all the files in our directory, we put all the arrays in one big array
• then we created an array for labels
• the shape of data fed to the model was 1280x19
• total no. of epoches =1565 , we splited them to 80% training , 10% valitation, 10% test
• then we scaled the data using standard scaler to ensure the data points as a balanced scale

1D CNN model

Model building:

• our CNN model consists of 15 layers
• the model consisted:
  • 6 convolution 1D layers
  • 6 pooling layers(2 max pooling and 4 average pooling)
  • 7 LeakyReLU layers
  • 2 dropout layers
  • 3 dense layers
  • The output layer is a Dense layer with a single unit and sigmoid activation for binary classification
• Optimizer used: Adam optimizer

Fifth: Confusion matrix

Sixth: Metrics

• We calculated the sensitivity, specificity, Precision, Recall, F-score and the test acuracy
  • Sensitivity: 90.36%
  • Specificity: 81.08%
  • Precision= 84.26%
  • Recall= 90.36%
  • F-score= 87.2%
  • Test acuracy= 86%

Seventh: ROC curve