Imperial College Institute of Biomedical Engineering: Cutting edge diabetes technology research.
Peter Pesl, Dipl.-Ing(FH), PhD Student
Mohamed El-Sharkawy, MSc, PhD Student
Pantelis Georgiou, PhD, Lecturer
Pau Herrero, PhD, Research Associate
Thanks to Aisling O’Kane for introducing me to the Imperial College Institute of Biomedical Engineering team. Peter Pesl showed me their Advanced Bolus Calculator for Diabetes (ABC4D), running on his iPhone. This prototype app will hopefully offer true smartphone based insulin dosage decision support. Unlike some other attempts, this app uses machine learning to refine suggestions based on past results. I was very impressed with this approach, and gave it a quick try. One enters current BG value, and planned carb intake. One then gets a dosage suggestion, which one can mark as accepted or declined. If declined, one inputs what you actually chose to take, and then later enter the results. Based on this record, the system learns your individual patterns, and improves its suggestions next time. For example, if you estimate carbs wrong, as long as you are consistent, the system will adjust.
In my conversations with diabetes app users, one of the big complaints was that the current apps didn’t really do enough to warrant the time invested. Just having a place to record your info doesn’t add a whole lot of value. An app that learns from your personal experiences and patterns, and then suggests how much insulin to take in a specific situation, seems like a great feature.
They have just finished their first round of clinical trials, hopefully it will prove safe and effective. I look forward to them either releasing it as a product, or allowing other apps to license the algorithm.
The following week in London, we had the opportunity to visit team leader Dr Pantelis Georgiou and the lab at the university. As well as the insulin dosage app mentioned above, the group is also working on a closed-loop insulin delivery system, which hopes to more closely mimic biological insulin release. This algorithm is placed on a small chip, which could be installed in pumps and is compatible with diverse systems.
This would be hybrid system, with some user interaction. The user would need to notify the pump that they are getting ready to eat, and size of meal. This would then cause the pump to give an initial insulin bolus. While this system would not be fully automatic, it could be very useful for many. This approach appears necessary at this time, as systems have struggled with the high BG values caused by the discrepancy between the time blood sugar rise is detectable, and the time it takes for the insulin to peak.
At the ATTD 2015, Novo showed data on a new faster insulin, which is quite likely attempting to address this issue.
For now, this hybrid approach will hopefully provide a work around for this particular challenge.
Was great to be able to be able to visit the lab and see these two projects. Inspiring work by a small but very dedicated team.
Thanks again for letting us to visit, and learn more about your work!
Please check out their work at the links below.