Researching brainwave patterns requires working with large data sets. This post explains how we can quickly review a huge CSV file generated from the MUSE EEG headband using simple command line tools. Muse has recently released a new app for iOS and Windows 10 called Muse Direct which allows for easy recording, graphing and savings of EEG data to the their cloud. Once we have a recording we can download a CSV file and summarize the data.
After launching Muse Direct and doing a brief recording we can see a new file is available on the Muse Direct Cloud website. We can download five different formats, but today we will focus on just downloaded the CSV file as it requires the least amount of tools to work with. The other formats have many dependencies on python libraries and Matlab (which starts at $150).
We will be using the csvkit to rapidly shape arrays into smaller datasets. This utility can be installed from the command line using:
$ pip install csvkit
Let’s get familiar with the data by checking out the first line of the file which names all the columns.
$ head -1 eeg.meditation.csv
timestamps,eeg_1,eeg_2,eeg_3,eeg_4,eeg_5,eeg_6,acc_1,acc_2,acc_3,gyro_1,gyro_2,gyro_3,batt_1,batt_2,batt_3,batt_4,drlref_1,drlref_2,mag_1,mag_2,mag_3,uv_1,uv_2,uv_3,press,temp,notch_filtered_eeg_1,notch_filtered_eeg_2,notch_filtered_eeg_3,notch_filtered_eeg_4,variance_eeg_1,variance_eeg_2,variance_eeg_3,variance_eeg_4,variance_notch_filtered_eeg_1,variance_notch_filtered_eeg_2,variance_notch_filtered_eeg_3,variance_notch_filtered_eeg_4,delta_absolute_1,delta_absolute_2,delta_absolute_3,delta_absolute_4,theta_absolute_1,theta_absolute_2,theta_absolute_3,theta_absolute_4,alpha_absolute_1,alpha_absolute_2,alpha_absolute_3,alpha_absolute_4,beta_absolute_1,beta_absolute_2,beta_absolute_3,beta_absolute_4,gamma_absolute_1,gamma_absolute_2,gamma_absolute_3,gamma_absolute_4,delta_relative_1,delta_relative_2,delta_relative_3,delta_relative_4,theta_relative_1,theta_relative_2,theta_relative_3,theta_relative_4,alpha_relative_1,alpha_relative_2,alpha_relative_3,alpha_relative_4,beta_relative_1,beta_relative_2,beta_relative_3,beta_relative_4,gamma_relative_1,gamma_relative_2,gamma_relative_3,gamma_relative_4,delta_session_score_1,delta_session_score_2,delta_session_score_3,delta_session_score_4,theta_session_score_1,theta_session_score_2,theta_session_score_3,theta_session_score_4,alpha_session_score_1,alpha_session_score_2,alpha_session_score_3,alpha_session_score_4,beta_session_score_1,beta_session_score_2,beta_session_score_3,beta_session_score_4,gamma_session_score_1,gamma_session_score_2,gamma_session_score_3,gamma_session_score_4,blink,jaw_clench,headband_on,is_good_1,is_good_2,is_good_3,is_good_4,hsi_precision,note,recorder_info,config,recording_notes
The csvstat included with the csvkit utilities provides a powerful summary for each column complete with# ranges for the values. Since this five minute recording resulted in a 85MB file let’s limit our query to looking at the first 500 lines of the file.
$ head -500 eeg.meditation.csv | csvstat
1. “timestamps”
Type of data:Number
Contains null values:False
Unique values: 106
Smallest value:1,528,057,484.807
Largest value: 1,528,057,486.13
Sum: 762,500,685,266.869
Mean:1,528,057,485.505
Median:1,528,057,485.49
StDev: 0.36
Most common values:1,528,057,485.17 (24x)
1,528,057,485.27 (23x)
1,528,057,485.78 (23x)
1,528,057,485.88 (23x)
1,528,057,484.96 (21x)
2. “eeg_1”
Type of data:Boolean
Contains null values:True (excluded from calculations)
Unique values: 1
Most common values:None (499x)
3. “eeg_2”
Type of data:Boolean
Contains null values:True (excluded from calculations)
Unique values: 1
Most common values:None (499x)
One fun stat that EEGs are good at picking up is eye blinking. We can get a summary of how many times I blinked my eyes during this session by using the following commands. It looks as though I blinked 27x!
$ head -1 eeg.meditation.csv | csvstat | grep -i blink
100. “blink”
$ csvcut -c 100 eeg.meditation.csv | csvstat
1. “blink”
Type of data:Boolean
Contains null values:True (excluded from calculations)
Unique values: 3
Most common values:None (499928x)
False (3446x)
True (27x)
Row count: 503401
Was I clenching my jaw? That would show up too.
$ head -1 eeg.meditation.csv | csvstat | grep -i jaw_clench
101. “jaw_clench”
$ csvcut -c 101 eeg.meditation.csv | csvstat
1. “jaw_clench”
Type of data:Boolean
Contains null values:True (excluded from calculations)
Unique values: 3
Most common values:None (499928x)
False (3384x)
True (89x)
Row count: 503401
The absolute readings for each brainwave {delta, theta, alpha, beta and gamma} on all four electrodes is interesting. We can get a quick summary for every brainwave on each electrode to narrow down what part of our head different brainwaves map to and how strong on average they are.
$ head -1 eeg.meditation.csv | csvstat | grep -i absolute
40. “delta_absolute_1”
41. “delta_absolute_2”
42. “delta_absolute_3”
43. “delta_absolute_4”
44. “theta_absolute_1”
45. “theta_absolute_2”
46. “theta_absolute_3”
47. “theta_absolute_4”
48. “alpha_absolute_1”
49. “alpha_absolute_2”
50. “alpha_absolute_3”
51. “alpha_absolute_4”
52. “beta_absolute_1”
53. “beta_absolute_2”
54. “beta_absolute_3”
55. “beta_absolute_4”
56. “gamma_absolute_1”
57. “gamma_absolute_2”
58. “gamma_absolute_3”
59. “gamma_absolute_4”
It might be easier to work with a seperate file so we can create eeg.absolute.csv which only contains the columns we are interested in examining.
$ csvcut -c 40-59 eeg.meditation.csv > eeg.absolute.csv
$ csvstat –mean eeg.absolute.csv
1. delta_absolute_1: 0.231
2. delta_absolute_2: 0.384
3. delta_absolute_3: 0.175
4. delta_absolute_4: 0.213
5. theta_absolute_1: 0.201
6. theta_absolute_2: 0.028
7. theta_absolute_3: -0.069
8. theta_absolute_4: 0.087
9. alpha_absolute_1: 0.796
10. alpha_absolute_2: 0.075
11. alpha_absolute_3: 0.174
12. alpha_absolute_4: 0.704
13. beta_absolute_1: 0.421
14. beta_absolute_2: -0.235
15. beta_absolute_3: 0.042
16. beta_absolute_4: 0.494
17. gamma_absolute_1: 0.219
18. gamma_absolute_2: -0.568
19. gamma_absolute_3: -0.046
20. gamma_absolute_4: 0.208
