Data Rep – Final Project

I want to use my personal diabetes data as the source for my final project. I have information from:

  1. 3 devices – insulin pump and two glucose monitors (data from two devices is available through a .csv file, the final device through a .xml file).
  2. Google doc – lists everything I have eaten this year
  3. Glycemic index – I found this spreadsheet of glycemic values posted on the blog of another diabetic @
  4. Photos – food images
  5. Exercise – I use RunKeeper and Gmap-pedometer to track my distances

I will start by focusing on the insulin and food data. If I have time, I can include the exercise information. But the medication and nutritional data is more important to me for this project.

With this, I will produce a Processing sketch or a series of sketches. In class this week, we saw several examples of great work that inspired me. The first was Ben Fry’s On the Origin of Species: The Preservation of Favoured Traces. As with many of Ben’s projects, it allows you to start with an overview and then drill down for more detailed information. I am particularly drawn to the animation. I have yet to build anything in Processing that uses the time function and works dynamically on this scale.

I was also inspired by Steve Varga’s Pennant iPad application, which was produced as his SVA thesis. It provides a thorough way of looking through baseball stats from many different directions.

The central issue with all this data is that it is hard to make sense of it all. I want to think about new ways of visualizing both the data and the relations between the different types of data so that people can learn more from it. I think the ideal system is one that stores this data and the right representation of that data in way that it is easily retrievable.

I expect that my representation will be best viewed on a large screen or in print. But if I have time, I’m also interested in experimenting with mobile device display. I think that would be most helpful for when people are eating at restaurants or shopping for food.

I plan on starting by visualizing my insulin data. The issue of latency is not addressed in any of the medical representations I have seen. Want I want to see is not when the dosages were taken, but how much insulin activity is happening right now. Perhaps the best way to do that is with a past, present and future view where a patient could use to adjust their medication/food balance.

The graph on the right show the activity level for my type of insulin (Humalog). On the left is an extremely rough sketch about how the two types of dosages I take could be plotted to show combined efficacy over time.

In trying to find the glycemic index spreadsheet this week, I also learned about other terms including glycemic load and insulin index. So glycemic index categorizes a food on a scale of 0 to 100, determining whether the index is low, medium or high. This number tells you the degree to which the food causes a spike in blood sugars. Glycemic load takes the index number and factors in the quantity of food being eaten.

Insulin index is a newer way of trying to determine a food’s impact based on blood insulin levels rather than blood sugars levels. It is based on the idea that “some foods cause a disproportionate insulin response relative to their carbohydrate load,” like certain proteins for example. I have yet to hear someone talk about insulin index and there isn’t much data available about it, so I can’t incorporate that into this project.

I will need to decide whether to allow live access to my data or aggregate it before I start. Aggregation is difficult between devices because of proprietary software that looks like it was built in the 80s. Some of it only works on a PC desktop and doesn’t have a web interface. I have been working with a developer to aggregate all this data in a MongoDB database. But I haven’t worked out pulling data from it. But this issue is certainly addressable in some form.