Hi all,
Long time fan, first time poster. I want to try and understand my physiology as a network, and to eventually be able to find network features that change with illness/seasons/events. Anyone interested in network modeling, data visualization, biomedical research, please comment and let’s see if we can work out some deeper/cooler stuff together.
I’m starting on a project to record as many physiological variables as I reasonably can with the following conditions:
continuous data (at least 1/15 min; more frequent is better).
data are free to access (device costs aside, but no subscriptions/services).
data are non or minimally invasive to gather (no implants, but will be using CGMs).
can be gathered longitudinally (at least over 1 week, ideally longer).
After each week of data collection, I will then use these data to explore network relationships between different outputs as functions of time, with an eye towards frequency composition. For example, during sleep, can I see sleep cycles across outputs, and if so, how much information could one output predict of another, and how much novel information does each output provide. Data will be supplemented with logs of food/sleep/wake logs as labels to find patterns against.
Outputs so far covered:
EEG (night only) - homemade electronics
EGG - homemade electronics
HR - homemade electronics, Oura ring
HRv - homemade electronics
“sleep” from Oura ring - Oura ring
Actigraphy - homemade electronics, Oura ring
Glucose - Freestyle Libre
Distal temperature - thermochron ibutton at wrist
Axial temperature - thermochron ibutton at arm pit
Let me know what I missed that you’re aware of/excited about.
Thank you!
I’ve added devices above. “Home made electronics” refers to a device built by a collaborator for research purposes. It’s collecting electrical activity from skin electrodes over the stomach for EGG, EKG (which yields HR, HRv), and I have extensions running electrodes to my forehead at night for EEG. It also has 3-axis accelerometers, so I can assess physical activity, but also should be able to reconstruct posture and rolling during sleep as well.
As for data management, I just keep it all local. Usually I convert time series data into 1/min per modality, and align them to some starting time, like midnight before the day I started recording. That way they then all line up as a matrix of output x minute, which is easier to plot/analyze. I do analysis in excel and matlab, and occasionally python (shout out to jupyter notebooks).
Fascinated by your project and your research. As a quantified-selfer on the consumer-side of data collection, I’m very interested in the data viz of my own personal data. However, it’s not easy to generate without having a comp sci degree, or an enterprise-level software subscription. Rich visualization that turns insight into Art, and wellness habits into easily identifiable patterns may intrinsinctly motivate behavior change in ways that points, leaderboards and badges may not. My point may be beyond the scope of this particular network physiology project but it would be cool to see physiology time-stamped in data viz, to serve almost like a digital Zeitgeber to counteract the feeling of the 24/7 culture
Interesting! Sounds like a cool project! I do a fair bit of network analysis in my work, and just recently started tracking my own physio data - it would be cool to think about using a network approach to understand it.
What are you thinking for the network modeling? I could imagine a lot of different possible ways to model it (what would be the nodes and edges; having different time points as different network layers for a multi-layer temporal network; calculating measures of network structure (or node-level measures that represent body systems) at different time points to compare; modeling how changes to one node cascade to changes in others with a transmission model, etc).