Welcome to the Artificial Pancreas Algorithm Project
The wiki about building a better pancreas that anyone can edit
35 articles since September 2007
About this project
The Artificial Pancreas Algorithm project is seeking people highly motivated to control an insulin pump with blood glucose data from a continuous blood glucose sensor. It's being studied elsewhere, but our algorithm will be different: we too will "close the loop" to control the basal rate of the insulin pump and drive the blood glucose level toward a target value, but we will additionally use state-based behavior to know when the closed-loop approach will perform better than the existing predictive approach and when it will not. We will also use adaptive variables and CGMS feedback for the purpose of "learning" the basal rates and C/I ratios and insulin sensitivity factors unique to each pump user. With a balance of the closed-loop approach and the predictive approach, we get the most out of the insulin pumps and CGMS sensors that we already have. It will be like having your own expert CDE built into your pump, and we can do this right now.
Motivation for an Artificial Pancreas
- A Patient's Story
- A Parent's Story
- A Caregiver's Story
Seeking New Members
Background Topics
- Endocrine physiology
- Insulin pump therapy
- Development of CGMS
- The "closed-loop" concept for an Artificial Pancreas
- A story problem just for fun
- Insulin pump control variables
- Expert Control
- Feedback of real-time blood glucose data to an insulin pump for bolus control
- Insulin and amylin combination
- Glucagon combination
- Further reading
Articles Explaining How We Can Close the Loop Right Now
- What does it REALLY MEAN to close the loop from the CGMS to the insulin pump?
- "Strike the Spike:" or THIS IS WHY the PID algorithm will never outperform a good prediction combined with a pre-bolus and closed loop correction
- Optional reading: Help with these funny bubble diagrams
- How will states and modes be used to control an insulin pump?
- Adaptive Variables, Adaptive Filtering
- Example: how will the algorithm "Learn" the basal rate?
- The JDRF Computer Simulation Tool for Algorithm Development
Helping out
- We need encouragement! Let us know you care by leaving a note on any author's "talk" page.
- Please rate an article you have looked at. It's anonymous, gives important feedback, and lets us know someone looked at an article.
- "Anyone Can Edit" includes you. If you see a spelling error, just fix it. Want to add a point? Just do it!
- Want to ask for more information, or make suggestions? You can do that on an authors talk pages or an articles discussion page.
To write a new article, just enter the article title in the box below.
<createbox> width=24 break=no buttonlabel=Create new article </createbox>
- Not sure where to start?
- Find out more about the wiki on the About page.
- If you are new to wikis, check out the tutorial.
- Check out Help:Starting this wiki if you're setting up the wiki.
- Adding content
- Every wiki has two list of articles that need help called "Stubs" and "Wanted Articles". Don't be shy, get in there.
- Uploading images is another really easy way to help out - see the Special:Upload page!
- You can find a list of useful templates on Category:Templates, some of which are documented on the templates project page.
- Talk and more...
- Check out the community portal to see what the community is working on, to give feedback or just to say hi.
Are you smarter than this 5th grader? Find out here ...
Living with Type 1 diabetes has its demands
The insulin pump delivers fast acting insulin precisely
The artificial pancreas combines the insulin pump with a continuous glucose monitoring system
Closing the loop: feedback of blood glucose values are used to adjust the insulin delivery
Existing insulin pumps use a predictive algorithm
The predictive algorithm works great if expertly and frequently calibrated
Adaptive Filtering: the artificial pancreas can learn your insulin needs, keeping your insulin pump experty calibrated
Given the 15 minute delay to deliver subcutaneous insulin to the blood stream, the PID control scheme will never control post-prandial glucose as well as a pre-bolus of insulin (typically 10-15 minutes before the meal) from the predictive model
State machine for the artificial pancreas algorithm controls the insulin pump appropriately: sometimes it is best to close the loop on basal rate (at night) but at other times it is best to close the loop on the predictive approach and adjust the insulin delivery for future use (after meals and corrections)