Autobetes is an app that can help to improve diabetes therapy. We develop algorithms to stabilise the diabetic’s sugar level. Ultimately, the algorithms auto-control the insulin pump based on coupled glucose sensor and minimal user input. Autobetes should minimise hypo and hyper health damage, reduce stress, hospital visits and health care costs, and make your life easier!

Our progress

The figure below shows our progress and plans. Our current app (1) collects relevant data, which we use to suggest how to (2) better configure your insulin pump in a TomTom-like fashion, and eventually (3) enable (semi) automatic pancreas software that takes over the insulin administration in most cases, just like a self-driving car!


(1) AutobetesQS: easy self-tracking in life sciences

Our Autobetes app enables easy self-tracking. It uses the accelerometer in your phone to auto-quantify your calorie usage when you start moving (walk, run, bike). It contains an extensive food list which makes it also easy to track your calorie and carb intake. You can easily store new activities, food and other events so you can mark them push-button if they’d happen again in the future.

Currently, type 1 diabetics can use Autobetes app to self-track all Quantified Self (QS) data that is relevant for their insulin therapy, including:

  1. Food (carbs)
  2. Physical activity (automatically estimated)
  3. Stress
  4. Blood and sensor glucose
  5. Insulin administration

We analyse the QS data 1-5 and provide personalised medicine/health feedback to the user (next section). We envision that in the near future Life Sciences will use/need QS data (and QS-tracking apps like Autobetes) more frequently to generate personalised medicine advice.

(2) Auto-config of the insulin pump as a TomTom

We have implemented our first statistical methods and models to analyse the QS data collected by users of the Autobetes app. While improved methods are underway, we already have our first prototype that automatically derives the crucial insulin pump settings from the data you collect:

  1. your insulin sensitivity
  2. your basal insulin needs
  3. your carb-ratio (i.e. the insulin you need to absorb the carbohydrates you eat)

Obviously, these settings may vary during the day according to biorhythm and recurring activity pattern. Therefore, we may advise different settings for different periods of the day. Moreover we want to show you the exact impact that physical activity has on your levels and how to best anticipate your insulin needs.

Our research aims to improve the insulin bolus calculator that is currently shipped with (Medtronic) insulin pumps. Additionally, we want to improve insight in the impact of your foods and activities on your levels.

(3) Autobetes pancreas software as a self-driving car

We also work on methods and models that may eventually (almost?) fully automate the control of your insulin pump. Briefly, we model the different organs in your body and continuously estimate the amount of insulin that is currently active, and e.g. the amount of carbs in e.g. stomach and guts. Doing so enables a prediction of the sugar level up to 2.5 hours in the future. Just enough to predict the right amount of insulin needed for the coming period. Discrepancies between predicted and observed sugar levels will lead to statistical optimal updates of all organs in the model. This way, the system may found out that you are eating or physically active – even if you have not informed it beforehand.

Technical overview

The figure below gives a technical overview of components involved in our project. Only thing you need, after installing our software on your phone and/or tablet, is marking the food you eat. Physical activities will be recorded automatically. Every now and then you upload your pump and sensor data to our server, after which you’ll find updated advices for basal insulin settings, insulin sensitivity and carb ratio in your app.


*Some of these components are still under development and only used and tested by a small group of selected patients, yet.

Of course, all of this needs careful and extensive testing, as we do not want you to suffer from sub-optimal advices. Therefore, you are more than invited to participate in our test panel and/or drop an email so we can keep you up-to-date on newest developments!