I built a smarter, lower-cost insulin pump with more advanced features than current pumps; it automatically dispenses insulin in response to blood sugar changes, as opposed to the current time-intensive process that requires diabetics to manually change pump settings multiple times a day. My pump costs $40 while current insulin pumps cost $6,000 – $10,000. I created the first pump-phone interface, allowing users to control the pump with their phones and receive text alerts of situations requiring intervention. Unlike current proprietary pumps, it works with any continuous blood-glucose monitor (CGM) and a wider range of insulin concentrations, making it more user-flexible.

I used a stepper motor controlled by a Raspberry Pi with a self-designed gearbox to enhance control of infusion rate; various force sensitive resistors to detect malfunctioning parts. My software improves on commercially-available pumps by adding a novel “closed-loop” that can automatically adjust insulin rates in response to blood sugar. To test the accuracy of the mechanics, I had the pump dispense liquid and weighted it using a precise scale. I user-tested by having diabetics interface with the prototype pump in parallel to their own. This testing indicates the pump is accurate plus/minus 3%, and this accuracy is similar to that of commercial insulin pumps, indicating potential applications for diabetes management.

 

What inspired you (or your team)?

The idea for my research started through conversations I had with diabetic friends. I heard their complaints about how their insulin pumps could only dispense insulin, and could do nothing to prevent low blood glucose, despite the fact that most diabetics wear devices to continuously measure blood sugar levels and the calculations to predict future blood glucose levels are known. This is particularly dangerous for them at night, as one of my friends is a deep sleeper and often does not wake in response to low blood glucose alarms. Initially, I wanted to make a system that would bridge the pump and glucose monitor data so that these lows would be prevented from occurring through suspension of insulin delivery. As my project progressed, my goal shifted from just preventing lows at night to maintaining normal blood glucose levels 24/7. My background research on other similar systems was done through a literature search and commercial search; I noticed that any similar systems in the pipeline suffered from issues with accuracy, cost, incompatibility, data visualization, and dependence on internet access. I hoped to make a safer, more user-friendly system that diabetics could actually afford to purchase.