Nexious is a project I started with Med after I stopped interning there. It's a project that I have been working/thinking about for the last few years, and something I got inspired while I was climbing. It's basically a way to use some of the research I did at Med on using decision trees to filter the probability of an individual having a particular disease with some data I logged while climbing. My idea was basically to find a relationship that would link the probability of having certain diseases to the altitude the individual is currently at (and a number of other factors).
This spawned from my love of mountaineering, and my passion to build things that would do some good. It would be made into a device that would use blood pressure, altitude, time spent climbing, and other factors to filter a few common diseases/conditions an individual could be having while they are climbing. My research was very specific as the data set I was using (self generated / I got some of the training data myself) was extremely limited so my altitude is always above 3000m. Using the APIs I created at Med and the knowledge behind creating decision trees I was able to come up with a pretty extensive model for the basis of altitude vs condition.
This is where Med came in. I needed help to build out a watch/some kind of portable device that would communicate with Android/iOS or even just some sort of external device to send it information about the current altitude, temperature, and time spent climbing by the individual. There are 10s of factors that determine how a person becomes ill or shows symptoms - so the mobile app is important as it has a series of questions the individual must respond to in order to reduce the error of the model. For example, the question could be "How is your stomach?", or "Do you have a headache?", and this would help us reduce some of the conditions that the individual could be showing. (This can also be diseases that could be happening simultaneous to the condition, but we account for that.)
We have many different prototypes for the design of the device, and we haven't gotten to a stage where we are trying to reduce costs. For the blood pressure we are using Pulse Sensor to test our values before we move ahead with the project. The toughest part has been to improve efficiency of the code so it can be used for a long-term (duration of climbing), and small in memory so it could be powered on any phone without cell service (pretty much the Internet).
I will be putting up more information as we progress, but this isn't going to be used in any commercial grade hardware/software. It's a very rough prototype as we don't have enough data to reduce error in our model. We have ran a lot of simulation/data analysis on the decision tree in our model, and we get a fairly small error compared to how much data we have. Our next steps would be to create devices that record these things, and give them to individuals who are climbing at high altitudes so we can use that data to make our model more accurate.
I am also no longer a major contributor in the project. I am still contributing code and some vision, but people who are better fit to generating data, and concentrate on the hardware in more detail have started to take this project more seriously. The most important part here is the data analysis and to see if we can actually create a good model for altitude vs condition.
Learning stack:
- Arduino, Java, Processing
- Started learning some very cool things in Arduino
- Started learning some pretty basic hardware (primarily just addons to the Arduino)