Medtronic
Medtronic Medtronic Medtronic Medtronic

What Control Means to Engineers

By Lane Desborough

Product Strategist

Posted:  7/28/2011 12:00 AM

Tags:

As many of you know, I came to work for Medtronic after my son Hayden was diagnosed with type 1 diabetes in August 2009. I was hired to apply my 25 years of experience as a control engineer for oil refineries, chemical plants, and energy companies to the task of developing an artificial pancreas. The work is challenging and fulfilling; it’s the most important work I’ve ever done. It’s also given me the chance to meet many other individuals and families faced with managing diabetes.

For example, earlier this year I had the opportunity to speak at Medtronic’s first Diabetes Advocate Forum. I really enjoyed the experience, connecting with so many powerful advocates for people with diabetes. So when the editors of The LOOP asked me if I wanted to talk to the DOC more often by blogging, I happily said yes.

To kick things off, I thought I’d talk about “control”. The word is used a lot in the diabetes community, often alongside labels like “good” or “bad”. But for engineers like me, it means something different. “Control” is an important element to consider as we work to develop an artificial pancreas. Check out the video below to learn more and let me know your thoughts in the comments.

This post is my July entry in the DSMA Blog Carnival. If you’d like to participate too, you can get all of the information at http://diabetessocmed.com/2011/july-dsma-blog-carnival/.

Comments

Bennet

Posted on Fri Jul 29 01:09:43 GMT 2011

Hey Lane!

Great to see you again even if it is only a poor online facsimile of the real you. Thanks for making this little video. I like the idea of thinking about control not as a bad word used to beat up people who struggle with their blood sugars but as a transfer of variability.

All the best.

Welcome to the diabetes blogging community.

Bennet

Karen G

Posted on Fri Jul 29 01:32:12 GMT 2011

Thanks for the insights - it's interesting to hear the point of view from someone on the other side of the technology coin!

Kelly Rawlings

Posted on Fri Jul 29 02:46:14 GMT 2011

Lane, me and my laggard pancreas really appreciate having you on the job of developing an artificial pancreas. Thanks for a great explanation of the challenges behind taking the concept to a real tool I hope to enjoy in my lifetime.

Cherise

Posted on Fri Jul 29 15:25:11 GMT 2011

Lane-

Thank you for sharing your perspective on diabetes technology and participating in the DSMA blog carnival.

Michael Hoskins

Posted on Sun Jul 31 02:12:31 GMT 2011

Lane: Thank you for this valuable insight on "control." Like the others who've commented, I too like going with "managing" rather than "control." Very interesting to hear you talk about that from the role of an engineer.

Scott K. Johnson

Posted on Tue Aug 02 21:15:51 GMT 2011

This is FANTASTIC.

I'm so glad that the rest of the world can get a peek at Lane's brilliance. Hope this becomes a very regular series.

Thank you!

David Humphrey

Posted on Wed Aug 03 16:29:50 GMT 2011

Lane, I am interested in more details of the variables in glucose control that you must deal with in designing the algorthmic control. Things like increased delay in glucose uptake into the cells as the recent (up to a day) caloric intake rises, or as the glucose level gets into the 300 range and above (which I hope the algorithms will prevent altogether. Other factors are illnesses, heavy exercise, etc. How will you tackle all these things?

Karrie

Posted on Thu Aug 04 17:51:00 GMT 2011

Hi David - thanks for the questions. Lane is actually travelling, but he'll take a look at your questions when he gets back.

Uzma

Posted on Tue Aug 09 23:11:14 GMT 2011

I think most of us think that having an artificial pancreas (or at least a closed loop system) will eliminate the need for us to do any more busywork, but after watching this video I'm not so sure. Will we still have to count carbs? Right now, I avoid them like the plague, finding that I digest my food much faster than my Humalog can reach it. What about hypoglycemia? Will we finally get to exercise without freaking out about it? And, finally, what about skipping meals, fasting, etc? Can we have diabetes without living with diabetes?

Lane

Posted on Mon Aug 15 22:40:00 GMT 2011

I share your desire to eliminate the tasks associated with diabetes management. We didn't ask for them, and they can take a lot of work. Something which we find very important to consider during the design of the artificial pancreas, however, are the trade-offs between safety, efficacy, and ease-of-use. There are some tasks - burdensome as they may be - which are more safely and effectively performed by humans than by computers. Humans are good at some things - detection of the novel, pattern recognition - and computers are good at others - vigilance, performance of rote procedures in well defined situations. Humans know things about the future which a computer never will: I'm going to exercise in an hour / I'm going to eat in 20 minutes / I'm going to sleep now. Just like you'd never turn your car's cruise control on in stop-and-go city traffic, there are many situations where automation is not as safe and effective as human control, and the human must still do busywork. Our goal is to safely and effectively "outsource" as many tasks as possible to the automation, providing dependable glucose control AND increased lifestyle flexibility AND reduced therapy effort.

David Guffey

Posted on Tue Aug 16 16:50:48 GMT 2011

Lane,
Great to see you still in the trenches moving "us" toward closed loop control (industrial automation term--interesting tie-in with the blog name). If "we" can install constraints into the automation and even have a stop/alarm if the automation gets too far out in the weeds, then wouldn't we feel safe enough to actually set that cruise control in the city traffic. To continue your analogy, it would be adaptive cruise control with the radar that applies the brakes when getting too close to a barrier.

I think that any progress will be good--even if there still is tedium to deal with. One of my favorite sayings from the Japanese is "perfect is the enemy of the good." Even if the closed loop control just took care of the unknown causes of BG fluctuation then that would be great. Not all the independent variables are measurable (noise?) so having a system that would at least respond and keep BG within a 20-30 mg/dl range would be so much better than the occasional mysterious-300 mg/dl-for-no-apparent-reason measurements.

I am just so pleased that you are working on this issue my friend. As now I know with confidence it will be solved soon.

Fred Goldberg

Posted on Wed Aug 17 15:32:06 GMT 2011

I think the video gave good background to what you are really doing. However as a T1 diabetic and Software Engineer I take issue with your comment that "diabetes is a simple single-input-single-output control problem" in Medtronic "News to Infuse Newsletter." Diabetes is really much more complicated I've learned diabetes is primarily a balance of Insulin, food, and exercise, with a little emotional variation thrown in. A serious control algorithm beside using exerciese must also look at rate of increase or decrease of blood sugar to offset some of the delay factor.

Lane

Posted on Wed Aug 17 17:33:00 GMT 2011

@David Great question! The design of control algorithms is a very mature engineering practice - at least in other domains - which means we can borrow heavily / stand on the shoulders of giants. Your question about delay, high glucose values, illness, and exercise highlight two of the many factors we're considering during the development of the artificial pancreas: unmeasured sources of variation, and system nonlinearities.

Illness, stress, exercise, and eating affect blood glucose. In control lingo we'd call these "unmeasured disturbances". Controllers - especially feedback controllers - are designed specifically for the case where the disturbances (sources of variation) are unmeasured (if we have measurements, then we can do a different kind of control: feedforward control). A common approach is to use "lumped parameter disturbance models".

Insulin resistance / insulin sensitivity (needing different amounts of insulin at different times / different levels of blood glucose in order to produce the same effect) is another important factor to consider. Control engineers call this "modeling the transfer function"; modeling the relationship between changes in the manipulated variable (insulin delivery) and the controlled variable (blood glucose). Sometimes the response is nonlinear (i.e. 1U of insulin may bring down blood glucose by different amounts); this must be factored into the design of the controller. This is what we'd call a nonlinear gain.

Lane

Posted on Fri Aug 19 21:08:00 GMT 2011

Thanks for your great comment. To clarify, my characterization of diabetes as a "simple single-input single-output control problem" was formed at my son's bedside in ICU the day he was diagnosed. By the next morning I'd figured out that it is far more complex than that :-). Although we can only "easily" measure one input (insulin) and one output (glucose), this doesn't mean this describes the system. Not by a long shot. This being said, there's a concept of "lumped parameter disturbances" wherein all the unmeasured disturbances are "lumped" together from a modeling perspective, allowing the design of a feedback controller to proceed. Sort of like the adage "worry about the things you can change", this is "worry about the things you can measure (and let feedback take care of the rest)".

I also agree with your point about rate of change of blood glucose. There is a wide and deep field of statistical techniques known as time series modeling. Characterizing dynamics - how something changes over time - is the fundamental objective of time series modeling. One of my favorite techniques is Box-Jenkins ARIMA (Autoregressive Integrated Moving Average) models but there are hundreds of different techniques, each with their own benefits and pitfalls.

Lane

Posted on Fri Aug 19 21:11:00 GMT 2011

It's so great to hear from you David! I think we learned a lot together about the human aspects of control when we worked together a lifetime ago. Like the human body, chemical plants are incredibly complex. But at the end of the day, humans are in charge of the automation of both systems, and it is incumbent on us as engineers to put the human in an "automated cockpit" where they can successfully accomplish their tasks.

Oh and as far as favorite Japanese sayings or words, how about these: go slow to go fast, Andon, Hansei, Yokoten, Obeya, Kanban, and Gemba!

Stay healthy my friend!

Lynda Griebrok

Posted on Sat Sep 17 01:49:17 GMT 2011

Why isn"t the medical field putting in real pancreas"s and using our personal stem cells. for rejection. Could this be an alternative to an artifical pancreas? I know that this has been done before but you must need 2 transplants. Why?

Erin

Posted on Mon Sep 19 22:04:00 GMT 2011

@Lynda There’s a lot of interesting work being done in biomedical research for diabetes. Kelly Close at diaTribe covers quite a bit of it (as well as medical technology). You might want to reach out to her/check out her site at http://www.diatribe.us.

Post a Comment

Thank you for posting on The LOOP!

We'll review your comment shortly. Please see our comment policy if you have any questions.

~The LOOP Team

Innovating for life.

x

You are leaving Loop-Blog.com

You just clicked a link to go to another website. If you continue, you will leave this site and go to a site run by someone else.

Medtronic does not review or control the content on the other website, and we are not responsible for any business dealings or transactions you have there. Your use of the other site is subject to the terms of use and privacy statement on that site.

If you want to visit the other website, click XXXXXX.com

If you do not want to visit the other website, click Cancel