TS2Image 🌠

This is the implementation of my work presented in partial fulfillment of the requirements for the degree of Bachelor in Computer Engineering: TS2Image: a software to convert EEG time series into images for training brain-computer interface convolutional neural networks

A tool to read EEG data from GDF files and export as images using Gramian Angular Field or ERSP.

Setup

Dependencies

TS2Image was developed using Python 3.8.2, that you can download and install from official site.

To install all third party dependencies run the following in your terminal:

pip -r install requirements.txt

Configuration

1. Set the directory containing the files you want to process:

input_folder = current_working_directory + "/datasets"

2. Set the root output folder:

output_folder = current_working_directory + '/images-output'

3. Create according to your dataset. For example, these are the events from some files in BCI IV competition dataset:

DESCRIPTION_EYES_OPEN = "276"
DESCRIPTION_EYES_CLOSED = "277"
DESCRIPTION_START_TRIAL = "768"
DESCRIPTION_CUE_LEFT = "769"
DESCRIPTION_CUE_RIGHT = "770"
DESCRIPTION_BCI_FEEDBACK = "781"
DESCRIPTION_CUE_UNKNOWN = "783"
DESCRIPTION_REJECTED_TRIAL = "1023"
DESCRIPTION_UNKNOWN_GROUP = "1072"
DESCRIPTION_EYE_MOVEMENT_HORIZONTAL = "1077"
DESCRIPTION_EYE_MOVEMENT_VERTICAL = "1078"
DESCRIPTION_EYE_ROTATION = "1079"
DESCRIPTION_EYE_BLINK = "1081"
DESCRIPTION_START_NEW_RUN = "32766"
BCI_competition_dataset_events_dictionary = {
    DESCRIPTION_EYES_OPEN:'Idling EEG (eyes open)',
    DESCRIPTION_EYES_CLOSED:'Idling EEG (eyes closed)',
    DESCRIPTION_START_TRIAL:'Start of a trial',
    DESCRIPTION_CUE_LEFT:'Cue onset left (class 1)',
    DESCRIPTION_CUE_RIGHT:'Cue onset right (class 2)',
    DESCRIPTION_BCI_FEEDBACK:'BCI feedback (continuous)',
    DESCRIPTION_CUE_UNKNOWN:'Cue unknown',
    DESCRIPTION_REJECTED_TRIAL:'Rejected trial',
    DESCRIPTION_UNKNOWN_GROUP:'Unkown Group',
    DESCRIPTION_EYE_MOVEMENT_HORIZONTAL:'Horizontal eye movement',
    DESCRIPTION_EYE_MOVEMENT_VERTICAL:'Vertical eye movement',
    DESCRIPTION_EYE_ROTATION:'Eye rotation',
    DESCRIPTION_EYE_BLINK:'Eye blinks',
    DESCRIPTION_START_NEW_RUN:'Start of a new run'
}

4. List of events from events_dictionary you want to export from your dataset:

valid_events_descriptions = [DESCRIPTION_CUE_LEFT, DESCRIPTION_CUE_RIGHT]

5. Start time window padding, in seconds. Negative values are accepted:

t_start = 0

6. Time window length, in seconds:

duration = 4

References

Encoding Time Series as Images for Visual Inspection and Classification Using Tiled Convolutional Neural Networks