EEG Based Depression Detection System

Files/Folders Info

1. EEG_128channels_resting_lanzhou_2015:

   • Format: .RAW files
   • 128-channel resting-state recordings
   • 24 Major Depressive Disorder subjects and 29 Healthy Control subjects
   • Age range: 16-52 years old
   • Includes: 1) demographic data, 2) psychological assessments.
   • Download Here

2. EEG_128channels_ERP_lanzhou_2015:

   • Format: .MAT files
   • 128-channel Event-Related Potential recordings
   • 24 Major Depressive Disorder subjects and 29 Healthy Control subjects
   • Age range: 16-52 years old;
   • Includes: 1) demographic data, 2) psychological assessments.
   • Download Here

3. SummarizedData:

   • ERP_Summary_limited.csv: Contains summarized data (16-electrodes) for ERP datasetfor all 55 patients.
   • Resting_Summary_limited.csv: Contains summarized data (16-electrodes) for resting dataset for all 55 patients.
   • Generated by running the Step1 notebook.

Step 1: Feature Extraction

1. Read the raw files from ERP and resting zips.
2. Visualizes the data.
3. Remove the noise using filters.
4. extract 16-electrode data from the 128-electrode dataset.
5. extract required features and unifies all 55 patient data into one dataframe.
6. convert the ERP and resting dataframes into .csv files for futher processing.
7. This step reduces the large MODMA dataset into very small sized filtered data for further processing!

• References
  • Collab file with explanation for Step 1
  • Description of the zip files. (just what the files contain, not much)
  • Freely available MODMA dataset

Step 2: Build ML Models

• References
  • Collab file with explanation for Step 2

References

1. Complete MODMA preprocessing and training for depression detection project
   • Full Article
   • Collab file with explanation for Step 1
   • Collab file with explanation for Step 2
   • Github repo for Step 2