The software uses support vector regression to find patterns in blood glucose levels over a period of time and predict future issues. In a similar fashion, support vector regression has been used in the past to predict stock market prices. "Support Vector Regression is a tool from machine learning that can build a regression model on the historical time series data in the purpose of predicting the future trend of the stock price." When graphed, blood sugars over time can look very similar to stock prices over time, so it is easy to see how the same technology can be applied.
 |
| Blood sugars from a few days on a graph |
From what I can understand, support vector regression works by using previous data and algorithms to generate new data. The system that is currently being developed would use factors such as previous blood glucose patterns as well as user input like "I'm going to go exercise now" or "I just ate a big dinner" and run algorithms to predict future blood glucose levels. After using this new form of software for a while, the computer will begin to learn how each individual person's body and blood sugar levels will react to different stimuli. The more a person uses this software, the more accurate predictions it will provide. This software would be incredibly useful and in some cases even life-saving for diabetics everywhere.
____________________________________________________________________________
References:
http://www.ourdiabetes.com/f/userfiles/image/dawn%20phenomenon.png (blood sugar picture)
http://graphs.net/wp-content/uploads/2009/03/stock-graph.jpg (stock picture)
http://ieeexplore.ieee.org/document/6703098/?denied (quote about SVR)
https://www.ohio.edu/research/communications/diabetes.cfm (article about blood sugar software)
Interesting post. Using computer science and technology for the good of the people is very important. This could change the lives of many people and possibly save a few lives. Personalizing the system to each person will give a level of accuracy to truly predict possible blood sugar spikes. This is done through computer science algorithms, and so is essential in saving lives.
ReplyDeleteIt's very fascinating how the data seems more regular than I thought. I thought it would vary due to differing times when one takes their insulin, when they eat, when they exercise, etc. such as the way the example graph shows the person's glucose levels reaching low points at consistent times over multiple days.
ReplyDeleteThe fact that glucose levels have a degree to regularity to them - aside from the deviations caused by eating, exercising... - makes this a prime piece of technology to develop so diabetics are less burdened by their condition and able to live better.
The part about accounting for things that may affect it is important too, a computer may be able to give specific advice for dosages based on the type of meal you're eating or how much you exert yourself in exercise.
Really cool post!
It is so awesome how developments are being made though computer science to help people. Not only are they creating this incredible technology, but they are also making something that is relatively complex accessible and easy to use. Although self driving cars and other things like that are really cool, I think it is so important that things like this continue to be developed to help people!
ReplyDeleteIt is so awesome how developments are being made though computer science to help people. Not only are they creating this incredible technology, but they are also making something that is relatively complex accessible and easy to use. Although self driving cars and other things like that are really cool, I think it is so important that things like this continue to be developed to help people!
ReplyDelete