Cortex data analysis pipeline for the LAMP Platform.

Overview

This API client is used to process and featurize data collected in LAMP. Visit our documentation for more information about using cortex and the LAMP API.

Jump to:

• Setting up Cortex
• Cortex example
• Find a bug?
• Developing Cortex
• Command-line usage

Setting up Cortex

You will need Python 3.8+ and pip installed in order to use Cortex. - You may need root permissions, using sudo. - Alternatively, to install locally, use pip --user. - If pip is not recognized as a command, use python3 -m pip.

If you meet the prerequisites, install Cortex:

sh pip install git+https://github.com/BIDMCDigitalPsychiatry/LAMP-cortex.git@master

If you do not have your environment variables set up you will need to perform the initial server credentials configuraton below:

python import os os.environ['LAMP_ACCESS_KEY'] = 'YOUR_EMAIL_ADDRESS' os.environ['LAMP_SECRET_KEY'] = 'YOUR_PASSWORD' os.environ['LAMP_SERVER_ADDRESS'] = 'YOUR_SERVER_ADDRESS'

Example: Passive data features from Cortex

The primary function of Cortex is to provide a set of features derived from pasive data. Data can be pulled either by calling Cortex functions directly, or by using the cortex.run() function to parse multiple participants or features simultaneously. For example, one feature of interest is screen_duration or the time spent with the phone "on".

First, we can pull this data using the Cortex function. Let's say we want to compute the amount of time spent by participant: "U1234567890" from 11/15/21 (epoch time: 1636952400000) to 11/30/21 (epoch time: 1638248400000) each day (resolution = miliseconds in a day = 86400000):

python import cortex screen_dur = cortex.secondary.screen_duration.screen_duration("U1234567890", start=1636952400000, end=1638248400000, resolution=86400000)

The output would look something like this: {'timestamp': 1636952400000, 'duration': 1296000000, 'resolution': 86400000, 'data': [{'timestamp': 1636952400000, 'value': 0.0}, {'timestamp': 1637038800000, 'value': 0.0}, {'timestamp': 1637125200000, 'value': 0.0}, {'timestamp': 1637211600000, 'value': 0.0}, {'timestamp': 1637298000000, 'value': 0.0}, {'timestamp': 1637384400000, 'value': 0.0}, {'timestamp': 1637470800000, 'value': 8425464}, {'timestamp': 1637557200000, 'value': 54589034}, {'timestamp': 1637643600000, 'value': 50200716}, {'timestamp': 1637730000000, 'value': 38500923}, {'timestamp': 1637816400000, 'value': 38872835}, {'timestamp': 1637902800000, 'value': 46796405}, {'timestamp': 1637989200000, 'value': 42115755}, {'timestamp': 1638075600000, 'value': 44383154}]} The 'data' in the dictionary holds the start timestamps (of each day from 11/15/21 to 11/29/21) and the screen duration for each of these days.

Second, we could have pulled this same data using the cortex.run function. Note that resolution is automatically set to a day in cortex.run. To invoke cortex.run, you must provide a specific ID or a list of IDs (only Researcher, Study, or Participant IDs are supported). Then, you specify the behavioral features to generate and extract. Once Cortex finishes running, you will be provided a dict where each key is the behavioral feature name, and the value is a dataframe. You can use this dataframe to save your output to a CSV file, for example, or continue data processing and visualization. This function call would look like this:

python import cortex screen_dur = cortex.run("U1234567890", ['screen_duration'], start=1636952400000, end=1638248400000) And the output might look like: {'screen_duration': id timestamp value 0 U1234567890 2021-11-15 05:00:00 0.0 1 U1234567890 2021-11-16 05:00:00 0.0 2 U1234567890 2021-11-17 05:00:00 0.0 3 U1234567890 2021-11-18 05:00:00 0.0 4 U1234567890 2021-11-19 05:00:00 0.0 5 U1234567890 2021-11-20 05:00:00 0.0 6 U1234567890 2021-11-21 05:00:00 8425464.0 7 U1234567890 2021-11-22 05:00:00 54589034.0 8 U1234567890 2021-11-23 05:00:00 50200716.0 9 U1234567890 2021-11-24 05:00:00 38500923.0 10 U1234567890 2021-11-25 05:00:00 38872835.0 11 U1234567890 2021-11-26 05:00:00 46796405.0 12 U1234567890 2021-11-27 05:00:00 42115755.0 13 U1234567890 2021-11-28 05:00:00 44383154.0} The output is the same as above, except the 'data' has been transformed into a Pandas DataFrame. Additionally, the dictionary is indexed by feature -- this way you can add to the list of features processed at once. Finally, a column "id" has been added so that multiple participants can be processed simultaneously.

Find a bug?

Our forum has many answers to common questions. If you find a bug, need help with working with Cortex, or have a suggestion for how the code can be improved please make a post on the forum.

Adding features to Cortex

If you are interesting in developing new features for Cortex, please check out our docs here. Note that the unittests in this repository will fail for users outside of BIDMC since you do not have access to our data.

Advanced Configuration

Ensure your server_address is set correctly. If using the default server, it will be api.lamp.digital. Keep your access_key (sometimes an email address) and secret_key (sometimes a password) private and do not share them with others. While you are able to set these parameters as arguments to the cortex executable, it is preferred to set them as session-wide environment variables. You can also run the script from the command line:

bash LAMP_SERVER_ADDRESS=api.lamp.digital LAMP_ACCESS_KEY=XXX LAMP_SECRET_KEY=XXX python3 -m \ cortex significant_locations \ --id=U26468383 \ --start=1583532346000 \ --end=1583618746000 \ --k_max=9

Or another example using the CLI arguments instead of environment variables (and outputting to a file):

bash python3 -m \ cortex --format=csv --server-address=api.lamp.digital --access-key=XXX --secret-key=XXX \ survey --id=U26468383 --start=1583532346000 --end=1583618746000 \ 2>/dev/null 1>./my_cortex_output.csv