ambient-bd-downloader 0.7.1

Download data from Somnofy devices

AmbientBD Somnofy data download

This a package for downloading sleep data from the radar devices and storing them in an "analysis friendly" format.

Installation

Setting up the environment

We recommend using Anaconda to manage your Python environment. Once you have Anaconda installed, you can create a new python environment by running the following command:

conda create -n ambient python=3.12

Activate the environment by running:

conda activate ambient

Installing the package

Install from the Python Package Index repository with the command:

pip install ambient_bd_downloader

Using the package

Remember to activate the environment first (conda activate ambient).

Configuration

On the first run, a properties file needs to be generated. To do this, run the command:

ambient_generate_config

This will create a file ambient_downloader.properties in the current directory.

Open the properties file with a text editor to change the parameters (description below).

Files and directories

client-id-file

Path to the the .txt file containing the client ID. The file must only contain the client ID.

Default value: ./client_id.txt

download-dir

Folder in which the data will be downloaded

Default value: ./downloaded_data

Data scope

from-date

The date from which to download data, in ISO format YYYY-MM-DD.

The program remembers the last downloaded session (for each study subject ) and continues download from there. Check below for more details on how to force the download of all sessions.

zone

The zone name(s) from which to download data. Multiple zones can be provided, separated by commas ,. zone=* will download data from all available zones.

device

The device id(s) from which to download data. Multiple devices can be provided, separated by commas ,. device=* will download data from all available devices.

subject

The subject identifier(s) from which to download data. Multiple subjects can be provided, separated by commas ,. subject=* will download data from all available subjects.

exclude-subjects

A pattern that can be used to exclude some subjects from the download based on their identifier.

• exclude-subjects=Test will exclude all subjects that have "Test" in their identifier (e.g. "Test ABD123" or "ABD123 Test")
• exclude-subjects=* will download all available subject (for the zone, device and subject lists specified)

Filtering

ignore-epoch-for-shorter-than-hours

Somnofy creates a new session everytime someone enters the room. These sessions are not useful for the analysis and can be filtered out from the epoch data. This parameter can be used to filter out the sessions shorter than the specified number of hours. These sessions will not be included in the epoch data but still visible in the session reports.

flag-short-sleep-for-less-than-hours

To help checking the compliance of the data capture process, the total hours of sleep are calculated for each night. If the total sleep time is less than the specified hours, the session is flagged as INVALID. This parameter can be used to set the threshold for the short sleep flag.

Download data

To download the data, navigate to the folder that contains the ambient_downloader.properties file, and run the following command:

ambient_download

Quality report

To generate quality reports on the data, navigate to the folder that contains ambient_downloader.properties and run:

ambient_quality_report

The quality reports use user-defined thresholds to flag sessions that might be of poor quality. These thresholds can be adjusted in the config file, under the [QUALITY_REPORT] section (if you do not have this section in your config file, you can get it by generating a fresh config file).

Quality Reports configuration

min-distance

The minimum average distance from the sensor over the session.

max-distance

The maximum average distance from the sensor over the session.

min-signal-quality

The minimum signal quality over the session (calculated from epoch data). A session will be flagged if > 80% of the epoch timepoints are below this threshold.

max-fraction-no-presence

The maximum fraction of "no presence" state during a session.

max-fraction-awake

The maximum fraction of "awake" state during a session.

min-session-separation

Minimum time interval (in minutes) between the end of a session and the start of the next. Any sessions separated by less than this threshold will be flagged as "split sessions" (the flag is applied only on the second session).

General behaviour

In the typical use, the program is run periodically (by a researcher or automatically) to download the data from the Somnofy server. The program remembers the last downloaded session for each subject and continues the download from there. For each run a new set of output files is generated and marked with the date range of the download (last_session date to current date).

It is recommended to run the program in the afternoon so all the "sleep" sessions will be terminated, as the program ignores the "in progress" sessions.

All raw data are stored in json files as retrieved from Somnofy.

The epoch data are the timeseries of the monitored variables, with time value in the timestamp column. They are called epoch because they contain aggregated data from the 30s epochs.

In the initial cleaning, the short sessions are filtered out from the epoch data. The short sessions are those which are shorter than the specified hours in the ignore-epoch-for-shorter-than-hours parameter in the ambient_downloader.properties file.

The compliance report is being generated for each night within the download date range. The compliance report contains for example the number of sessions, total sleep time and the flag VALID. The night is considered valid if the total sleep time is greater than the specified hours in the flag-short-sleep-for-less-than-hours parameter in the ambient_downloader.properties file.

Monitoring program porgress

The program outputs the current operations or errors to both the console and the log file. The log file is stored in the download directory as download.log.

Data layout and format

Example folder structure for 2 subjects (ABD001, ABD002) in zone "ABD Test Zone" after 2 download program runs, and one quality control run. on 2024-07-26 and 2024-08-03. There were no data for subject U03 after 26th July.

- downloaded_data
    - Session_qc_2024-07-14_2024-07-26.csv
	- Participant_qc_2024-07-14_2024-07-26.csv
    - ABD Test Zone
        - ABD001
            - data
                - 2024-07-14_2024-07-26_epoch_data.csv
                - 2024-07-14_2024-07-26_session_report.csv
                - 2024-07-14_2024-07-26_compliance_info.csv
            - raw
                - 2024-07-14_WUVFRRgHDg4RAgAA.json
                - 2024-07-14_WUVFRRgHDgkCIgAA.json
                - ...
            - sys
                - last_session.json
        - ABD002
            - data
                - 2024-07-14_2024-07-26_epoch_data.csv
                - 2024-07-14_2024-07-26_session_report.csv
                - 2024-07-14_2024-07-26_compliance_info.csv
                - 2024-07-27_2024-08-03_epoch_data.csv
                - 2024-07-27_2024-08-03_session_report.csv
                - 2024-07-27_2024-08-03_compliance_info.csv
            - raw
                - 2024-07-15_WUdIQhgHDxYHIgAA.json
                - ...
            - sys
                - last_session.json

For each subject a folder is created in the download folder. Inside the subject folder, there are three subfolders: data, raw and sys.

The sys folder contains the information used by the program to track the download status. For example it stores the last finished session which has been downloaded. This information is used when download is restarted to continue from the last session.

The raw folder contains the raw data downloaded from the Somnofy server. The data is stored in json format. Each session is stored in a separate file. The file name is the DATE-SESSION_ID.json. That way the raw data can be sorted by date.

The data folder contains the epoch data and sessions report in an easy to use form as data tables. The data is stored in csv format. Whenever the download is started, the date range is established as START_DATE_OF_THE_FIRST_DOWNLOADED_SESSION to the END_DATE_OF_THE_LAST_FINISHED session.

For example if in the previous run the last downloaded session terminates on 2024-07-14 at 9:30, we run download in the afternoon of the 2024-07-26 the program will use as the date range: 2024-07-14_2024-07-26 (if there was a session on the 26th July).

Additionally, the file all_sessions_report.csv contains characteristics of all the sessions ever downloaded by the program, which can be used to find sessions for any day within the study.

Epoch data

The epoch-data file contains the timeseries of the monitored variables for each session within the download date range (remember only the sessions with duration larger than ignore-epoch-for-shorter-than-hours are included).

The epoch data contains the following columns:

Column Name	Description
timestamp	The timestamp of the 30 second epoch for which data row is created.
session_id	The Somnofy's unique identifier for the session.
signal_quality_mean	The mean signal quality during the epoch.
respiration_rate_mean	The mean respiration rate during the epoch.
respiration_rate_var	The variance of the respiration rate during the epoch.
movement_fast_mean	The mean movement detected during the epoch.
movement_fast_nonzero_pct	The percentage of non-zero values for fast movement during the
heart_rate_mean	The mean heart rate during the epoch.
heart_rate_var	The variance of the heart rate during the epoch.
external_heart_rate_mean	The mean external heart rate during the epoch.
external_heart_rate_var	The variance of the external heart rate during the epoch.
external_spo2_mean	The mean external SpO2 level during the epoch.
external_spo2_var	The variance of the external SpO2 level during the epoch.
distance_mean	The mean distance to subject measured during the epoch.
motion data count	The number of data points in the epoch (30).
light_ambient_mean	The ambient light level during the epoch.
sound_amplitude_mean	The sound amplitude during the epoch.
temperature_ambient_mean	The ambient temperature during the epoch.
pressure_mean	The mean atmospheric pressure during the epoch.
humidity_mean	The mean humidity value during the epoch
indoor_air_quality_mean	The mean indoor air quality during the epoch
epoch_duration	The precise duration of the epoch (seconds)
sleep_stage	The sleep stage as calculated by the VitalThings algorithm. 1 = Deep sleep, 2 = Light sleep, 3 = REM, 4 = Awake, 5 = No presence

Compliance info

The compliance info file contains rows for each nights within the download date range with characteristics that help assess the validity of the sleep data, for example if the number of sessions is 0 the sensor must have been offline.

The compliance info contains the following columns:

Column Name	Description
night_date	The date of the night for which sessions are being assessed. All sessions which end at this date are included.
number_of_long_sessions	The number of sessions longer than the minimal duration specified in ignore-epoch-for-shorter-than-hours.
max_time_in_bed_h	The maximum time spent in bed during the night, in hours
max_time_asleep_h	The maximum time spent asleep during the night, in hours.
total_sleep_time_h	The total sleep time during the night, in hours. Calculated with time asleep for all not ignored session in this night
valid	True if the night is valid, which currently means: (a) there are recordings for this date, and (b) the total sleep is >= flag-short-sleep-for-less-than-hours

Session report

The session report file contains rows for ALL session within the download date range (they are not filtered with ignore).

Apart from housekeeping information about the session (id, session_start, session_end), it contains the information about the sleep and activity.

Column Name	Description
id	The somnofy unique identifier for the session.
state	Indicates if the session is ENDED or IN-PROGRESS
subject_id	The unique ID of the subject associated with this session.
device_serial_number	Serial number of the Somnofy device which recorded the session.
epoch_count	The number of 30s epochs in the session.
session_start	The timestamp for start of the session.
relative_session_start	@TODO check somnofy
session_end	The timestamp for end of the session.
time_at_last_epoch	The timestamp the last epoch of the session.
time_at_intended_sleep	The time of sleep attempt.
time_at_intended_wakeup	@TODO CHECK SOMNOFY
time_in_undefined	The time spent in undefined (sleep) stage.
time_at_sleep_onset	The fall asleep time.
time_at_wakeup	The wakeup time.
time_at_midsleep	Time at the mid-point between sleep onset and wakeup.
sleep_onset_latency	The time it took to fall asleep in seconds.
sleep_period	The duration of the sleep session [s].
time_in_bed	The total time spent in bed [s].
time_asleep	The total time scored as asleep in the session [s].
sleep_efficiency	The efficiency of sleep.
time_in_light_sleep	The time spent in light sleep [s].
time_in_rem_sleep	The time spent in REM sleep [s].
time_in_deep_sleep	The time spent in deep sleep [s].
time_in_no_presence	The time spent with no presence detected [s]
number_of_times_no_presence	The number of times no presence was detected.
time_wake_after_sleep_onset	The time spent awake after sleep onset.
number_of_times_awake	The number of times the subject woke up.
number_of_times_awake_long	The number of times the subject woke up for a long duration.
time_wake_pre_post_sleep	The time spent awake before and after sleep.
time_from_sleep_onset_to_first_rem	The period of REM sleep onset.
movement_fast_during_sleep_period_mean	The mean movement during sleep.
rpm_non_rem_filtered_mean	The mean RPM (respirations per minute) during non-REM sleep.
rpm_non_rem_mean_var	Variance of the RPM during non-REM sleep.
rpm_non_rem_baseline	Baseline of the RPM during non-REM sleep.
rpm_non_rem_baseline_std	Standard deviation of the baseline of the RPM during non-REM sleep.
heart_rate_non_rem_filtered_mean	The mean heart rate during non-REM sleep.
heart_rate_non_rem_mean	The non-filtered mean heart rate during non-REM sleep.
external_heart_rate_non_rem_filtered_mean	The mean external heart rate during non-REM sleep.
epochs_with_movement_pct	The percentage of epochs with movement.
epochs_with_movement_during_time_in_bed_pct	The percentage of epochs with movement while in bed.
time_with_movement_pct	The percentage of time with movement.
time_with_movement_during_time_in_bed_pct	The percentage of time in bed with movement.
sleep_cycle_count	Number of sleep cycles in the session.
is_workday	Whether the session was on a work day
chronotype	The chronotype for this session, i.e. the time at mid-sleep, in hours.
social_jet_lag	@TODO: CHECK SOMNOFY
jet_lag	@TODO: CHECK SOMNOFY
regularity	@TODO: CHECK SOMNOFY
sleep_score_standard	The sleep score based on the standard scoring method.
sleep_fragmentation	@TODO: CHECK SOMNOFY
external_spo2_mean	The mean external SpO2 level.
external_spo2_min	The minimum external SpO2 level.
distance_during_sleep_mean	Mean distance from the sensor during sleep.
distance_during_sleep_std	Standard deviation of the distance from sensor during sleep.
air_pressure_filtered_mean	The filtered mean air pressure.
light_ambient_filtered_mean	The filtered mean ambient light level.
light_ambient_at_wakeup_mean	The mean ambient light level at wake-up time.
sound_amplitude_filtered_mean	The filtered mean sound amplitude.
sound_amplitude_during_sleep_filtered_for_noise_mean	The mean sound amplitude during sleep, filtered for noise.
sound_amplitude_during_sleep_filtered_for_movement_mean	The mean sound amplitude during sleep, filtered for movement.
sound_amplitude_at_wake_up	The sound amplitude at wake-up time.
epochs_with_sound_spikes_during_sleep_count	Number of epochs that had sound spikes during sleep.
awakenings_caused_by_sound_count	Number of awakenings caused by sound.
temperature_filtered_mean	The filtered mean temperature.
indoor_air_quality_filtered_mean	The filtered mean indoor air quality.
air_humidity_filtered_mean	The filtered mean air humidity.

All sessions report

Is the same as session_report but contains all the sessions ever downloaded by the program. It can be used for example to find sessions for any day within the study.