Are you conscious? Now with our consciousness detector, you can find out!

Hello again! This is the mind-reader reporting to you with updates on my project. I have had quite the scientific adventure since last sharing my research so sit down, grab your tea (or coffee or pop or kool-aide – I don’t judge) and prepare for a rollercoaster.

With no success from LED oddball tasks, I moved to replicate an auditory oddball task from a paper that describes P300 responses from minimally conscious and vegetative subjects. If subjects with severe brain damage are able to produce results from that task, shouldn’t a healthy brain produce them as well? With this thinking in mind, I created a task that produces an Arduino-driven tone from a buzzer that lasts 100 ms, with 900 ms between each tone.  The oddball tone is coded to appear 14% of the time. When this tone appears, the subject makes a tally until we reach 50 tallies, as the P300 signal is reliable after 30 to 40 oddball stimuli have been presented. The signal sent to the buzzer is essentially copied and sent to the EEG so that the tone activation can be seen in the Spike Recorder app, as shown below.

With this information in the app, the data can be averaged around tone onset. I set out to make this work – except it didn’t. Trial after trial returned a flat average. I was finding something that I thought looked like the P300 but the absence of anything substantial from the average suggested that what I was looking at was not consistent enough to be called science.

This lack of success caused me to scale back the project and start from the absolute basics.

One of the current BYB EEG experiments involves finding the alpha wave: a 10 Hz signal that appears from the occipital lobe when the eyes are closed (can be viewed at https://backyardbrains.com/experiments/eeg). This experiment was used as a control to ensure that the EEG was working as it should. We attached three shields to the board to allow for three recording locations: occipital lobe, right temporal lobe, and a forehead control.

To ensure that activity was not dependent on the shield, we cycled the inputs from each recording location so that every location was recorded through each shield. The results confirmed that the alpha wave is most intense over the occipital lobe, less intense but still visible over another cortical location, and nonexistent over a non-cortical location (changes in intensity can be seen in RMS values). With confirmation that the shields are functioning as they should, I climbed to the next control: the flash visual evoked potential.

Flash visual evoked potentials (fVEP) represent electrical signals generated by the occipital region of the cortex when the subject is stimulated with flashes. The main components of the signal are those displayed to the right and are named for their latency, which is highly variable between subject and task, and their polarity. The flash task created to elicit this waveform was powered by an Arduino and a surge protector that has been engineered to receive power inputs through a wire. The Arduino sends constant power to the bulb until the push of a button begins light flashes at a rate of 1/sec for 60 ms each. Each recording begins with an alpha task to ensure that the signal is legitimate. After the signal is verified, the subject sits motionless in my office for one minute and watches the flashing of the bulb. Because of the small amplitude of the fVEP response, the waveform is easily
lost in a raw EEG signal. It is only through averaging of trials that this evoked potential is visible, since the information common to the entire recording will be averaged out. Errors in the Spike Recorder software averaging caused us to call in Matlab for offline data
analysis. One second of data was collected surrounding flash onset and all of these epochs were averaged after eliminating outlier responses. The fVEP mean is then plotted against a Monte Carlo mean to show where and when the data is statistically significant – any data falling within the 95% confidence interval is deemed insignificant. If the data is significant and the waveform components match the literature in latency and amplitude, I considered the trial a success. Several successful trials indicated to me that the fVEP procedure produced what was necessary for the signal to appear and that the data analysis allowed us to see this particular event-related potential. Hoorah! It is possible. Equipped with new Matlab skills and some inspiration, I refocused my project to finding the P300.

My initial set-up for the oddball task was not scientifically sound, so some adjustments to better control and record the stimuli were necessary. With a better designed experiment and several loyal subjects, data collection was in full swing. After collection, the data was run through an adapted Matlab script specific to the task. This script creates a plot of 1.4 seconds surrounding standard tone onset, 1.4 seconds surrounding oddball tone onset, a Monte Carlo simulation, and a plot of all three plotted together for comparison with outlier data excluded, same as before.

The code outputs the largest positive potential between 300 and 600 ms after tone onset, displaying the latency and change in amplitude from baseline for that point. The results are very exciting! We appear to have a P300 on our hands. Nearly half of the recordings taken thus far have had significant results. As I am only three days of data collection in, I’m happy with that! A lack of significance in the other trials could be from poor recording location, high impedance between the electrode and the skin, or simply poor attention allocation on the part of the subject. My goal now is to keep the positive results coming – more collection, more collection, more collection! Replication = science, right?