Recent years have shown an increasing interest in the development of neuro-feedback systems for intracranial EEG experiments and clinical applications. Such systems may, for example, allow stimulus manipulation in response to certain characteristics of the real-time neural signal. They are of great importance for testing the predictions about the functional role of neural oscillations of various frequencies found in EEG signals. In collaboration with the research group of Dr. L.M. Talamini (Brain and Cognition program, University of Amsterdam) we have developed a neurofeedback extension for EventIDE (our stimulus presentation software) that allows designing experiments, where stimulus presentation is accurately locked to a selected phase of the online EEG signal. Below, you can see a demo of EventIDE phase-locked stimulus presentation made over a real EEG recording.
|Real-time phase forecast prediction and stimulus presentation locked to the upper phase of slow oscillations (~1Hz) in an EEG signal. The noisy EEG signal is shown in cyan, whereas its filtered counterpart in white. The present time moment corresponds to the ‘zero time’ mark on the X-axis. Made phase forecasts are shown with the vertical orange bars. Presented stimuli (short sounds) are represented by the horizontal color strokes.
The extension consists of a hardware driver and signal-processing module. The hardware driver performs real-time acquisition of single-channel data from an EEG amplifier. Based on that data the signal-processing module continuously estimates a time interval to the next predicted occurrence of the selected phase in an online neural signal. The estimated time interval is then used by EventIDE to initiate a locked stimulus presentation.
The signal-processing module performs the following functions in each of continuous processing steps in a moving signal window (each step takes ~100 ms over a window of 5000 samples):
- Estimation of frequency power in the chosen range
- Instantaneous phase estimation
- Adaptive bandpass filtering of the signal
- Time-series forward prediction
- Real-time plotting of the signal window, frequency power and signal forward prediction
- Collection of numerical signal statistics, e.g. the current dominant frequency
- Estimation of the time interval to the next phase-locked stimulus presentation
- Evaluation of the past predictions in respect to the real signal
We tested the extension both in real EEG recordings and an emulation mode, in which a large amount of offline data was used. We found that the extension:
- makes accurate phase predictions even in a noisy and frequency unstable signal
- can make prediction for any signal phase (not only signal peaks)
- can be used for a wide range of oscillation rhythms after minimal tuning
- provides comprehensive and detailed statistics for real-time monitoring
- allows a researcher to control and adjust the estimation procedure in real-time
- allows presentation an arbitrary stimulus (e.g. sound, picture) locked to a signal phase with 1ms accuracy.
- can be easily adapted to the existing experimental setups
- demonstrates good performance on a standard PC
In sum, a combination of our stimulus presentation software and the neurofeedback extension offers a complete solution for designing experiments, where stimulus presentation is based on real-time neurofeedback characteristics, in particular, a selected phase of the EEG signal.
At the following stage we intend to improve accuracy of the phase prediction algorithm based on results obtained in a large-scale testing. We also plan to adapt the extension for forecasting in other types of neural signal, like MEG. Therefore, we are seeking for collaboration with research groups that are interested in applying neurofeedback methods and use our software.
Currently, the extension supports three popular classes of EEG amplifiers: Brain Vision, OpenEEG and TMSi/ANT. If you are interested to enable support for different amplifier models, available in your lab, please, contact us. In the most cases it will be an easy job on adaptation of your hardware.