edf-importer

Tool to import .edf files to influxdb or victoriametrics. I created this tool to import data from my ResMed AirSense 11 AutoSet CPAP machine into VictoriaMetrics. It may not necessarily correctly import data from other models or brands yet. It does support both timeseries metrics and annotations (events).

Caveats

• The path is expected to be a directory filled with datestamped dirs in YYYYmmdd format.
• The filenames to read are currently requiring to end in BRP.edf, EVE.edf, or PLD.edf.
• EDF / CPAP data is usually sampled at a very high rate. Make sure your database is configured to allow this. For Victoriametrics, for example, make sure the flag -dedup.minScrapeInterval is set to a low value, in my case less than 40ms.
• The data is read with your computer's local timezone, as EDF files do not necessarily have a timezone associated with their data.
• Zero-duration events from EDF+ annotations are automatically given a 10 second duration. This was done to allow easier graphing of ResMed Hypopnea events.

Running

Copy edf-importer.example.yaml to edf-importer.yaml and fill in your database config.

Edf-importer will continue to run and watch the specified path, importing new data the next time you insert your SD card.

./edf-importer --config edf-importer.yaml --path /Volumes/<drivename>/folder    

Usage of ./edf-importer:
  -config string
    	Config file (default "edf-importer.yaml")
  -dry-run
    	Don't insert into the database
  -path string
    	Path to data directory (default "/Volumes/NO NAME/DATALOG")
  -state-file string
    	State file (default "edf-importer.state.yaml")
  -v	Show version and exit

Running as a macOS LaunchAgent

• Create a directory for edf-importer to live in. I chose /Users/me/Documents/edf-importer.
• Copy to the directory:
  • The edf-importer binary
  • Download edf-importer.example.yaml, rename to edf-importer.yaml and fill in your values
  • Download com.github.tedpearson.edf-importer.plist.example, rename to com.github.tedpearson.edf-importer.plist, and modify the paths to point to the directory you chose above, and change the --path argument to point to your SD card volume.
  • Move com.github.tedpearson.edf-importer.plist to ~/Library/LaunchAgents
  • Start edf-importer with launchctl bootstrap gui/$(id -u) ~/Library/LaunchAgents/com.github.tedpearson.edf-importer.plist
  • Stop edf-importer with launchctl bootout gui/$(id -u) ~/Library/LaunchAgents/com.github.tedpearson.edf-importer.plist

Details

Info about exported data:

Metric name: cpap

• fields:
  • "edf supplied name"
    • name will be whatever the EDF label name is from the EDF file.
    • contains the raw data from the EDF file
  • annotation
    • value of 1 during the event
    • value of 0 before and after the event
• tags
  • event
    • contains the EDF+ annotation name
    • only on annotation events, not raw data

Dashboard

I have included my Grafana dashboard definition in the repo. I based it heavily on the graphs in OSCAR (see Other Resources).

Note that the dashboard as exported uses a VictoriaMetrics (Prometheus-like) datasource, not an InfluxDB datasource.

Features:

• Annotation/Events, large leak, and flow limit color block display, which are toggleable
• Graphs of Flow Rate, Pressure, AHI, Snore Index, Flow Limit index, Leak Rate, Respiration Rate, Minute Ventilation, and Tidal Volume.
• Last Night singlestat view. Reports midnight-midnight values only. Only valid when the end of your time range falls during an awake period.
• AHI/Event daily bar charts. Report midnight-midnight values only.

Here are a few screenshots of the dashboards:

• A week, currently the default view, showing off AHI/Events per day views
• An entire night
• A zoomed in view so you can really see the details of the annotations and individual breaths.

Dashboard - Week

Dashboard - Full Night

Dashboard - Zoomed

Other Resources

Edf-importer was heavily influence by and my dashboards were inspired by OSCAR, the Open Source CPAP Analysis Reporter. OSCAR allows you to import and view all the same data, see overviews and summary reports, and also has some additional calculated metrics including inpiration time, expiration time, and time at pressure. I will note that its AHI graph resets every time there is a data gap (when you take off the mask), so my dashboard is more accurate in this one case.

For many people, OSCAR is probably all you need. For the nerds, this project is for you.