AI + EEG + Meditation
Exploring Daniel Ingram's Meditation Style









                

There is growing interest in using EEG, the electrical signals from the brain, to explore the effects of meditation. Daniel Ingram is a meditation teacher with many years of experience practicing a traditional Theravadan Buddhist method often called the Stages of Insight. He generously offered to record his method using the Muse EEG, an affordable model our team is using to informally survey meditators. Daniel recorded seven meditation sessions over a one-month period, shown in this graph, one color for each session. This webpage exhibit explores some possible insights from this preliminary data. This is not peer-reviewed research and the sample sizes are small, but the very strong, replicable data shows great promise. We hope this exhibit helps educate other developers and meditators, and we thank the members from the EPRC who participated, including Daniel.



            

First, let's focus on just one meditation session. The red line in the chart shows a quickly increase gamma power on the front-left part of Daniel's brain. Higher "power" can be thought of as a higher "volume" of this frequency. The power increases massively, almost 500%, an unmistakable impact of the meditation on the brain. It is well known from many scientific studies that meditation often increases gamma power, so this result is in line with the best available science. However there is still a surprising lack of consensus about what exactly this increase means. Daniel's data suggests a strong correlation between the progress in his practice and this specific measurement. It's a useful indicator, and for now that's enough.



            

Averaging all seven meditation sessions together, the white line shows the average and the grey area shows the range of all recordings. The top of the grey area are the maximum values recorded from any meditation session, and the bottom shows the minimum values (the slowest gamma increase). What is clear from this chart is that every meditation session raises gamma power, within a certain dependable range.

This chart matters because science depends on repeatability. If just one recording showed a gamma increase, that could be due to randomness. The next recording might show a gamma decrease, or no change. The possiblity that seven recordings in a row shows the same behavior is very low. Any signal that is this dependable is useful - we can build software that is genuinely useful to meditators. At least, this analysis would be useful for Daniel, who it was designed for. What about other people?





            

Meditation participants Kaio and Niffe, practicing the same Insight methods as Daniel, have similar EEG results, with a consistent increase in gamma on the front-right part of the brain. These results show "replicability" - the important criteria in science which states that independent researcher should be able to reproduce an experiment under identical conditions and obtain the same results . If the EEG data from all meditators we recorded looked completely different, it would suggest that this method of analysis lacks general usefulness, perhaps because the hardware or software is insufficient for the goal.




            

In contrast, participants who practice different types of meditation see different gamma behavior. Two participants, Steffan (myself) and Doug, have consistently large increases in TP10 gamma (above the right ear), not AF8 gamma (right forehead). This difference was expected because our methods use intuitive visualizations or 'active imagination' - a completely different practice from the "Stage of Insight" that Daniel, Kaio, and Niffe practice. In fact, Daniel's TP10 gamma barely changes at all during his practice. Usually, it even decreases! These results suggest that only two electrodes are needed to make a simple classification of basic meditation style.





            

AI can be used to classify these meditation styles. This graph shows Daniel's first meditation session compared with other participants. Daniel matches his own style best (his line reaches 100% match). Kaio and Niffe have strong matches with each other and with Daniel because they practice the same method. Steffan and Doug have NEGATIVE matches with all other partipants (as low as -40% match) because their method is completely different. This simple analysis shows that standard AI methods can easily classify different styles - a useful result for new meditators who are unsure if their practice is "correct", or for teachers who want to track their students' progress.




            

Do you want to participate? It's easy! All you need is a Muse EEG and the Android/iPhone app Mind Monitor to record data. This data can be uploaded to my free website Meditation Monitor (your data is processed in the browser, not stored or sent to any servers). Here is a 10-minute YouTube tutorial for how to get started. Feel free to email me at the address below. for guidance or to request a custom analysis of your data. Also, if anyone wants to help improve these websites to make them more useful, please let me know.