The Personal Glucose Predictive Diabetes Advisor (DIAdvisorTM), is a FP7 project developed within the framework of a 14 partners consortium, including some prestigious universities, and companies in various fields of activity.


Functioning Principles

A simple logic that implies patient input, data interpretation, user friendly feedback, and information sent to doctor

Main features

The platform for the target implementation is a handheld device like a PDA or a mobile phone which the patient can carry anywhere during the day.

The flow through the BG (blood glucose) predictor system can be divided into three steps:

Inputs from vital sign sensors and blood glucose measurements are acquired from the patients;

A prediction engine consisting of a mixed model, combined with a predictive control algorithm, predicts the future blood glucose development;

A decision support module interprets the predicted BG curve


A personalized blood glucose predicting system, which can be used on the spot to assist patients struggling with diabetes, in various daily situations. The device can predict the blood sugar highs and lows with over 90% accuracy at least 20 minutes

The system includes a clinician application module,

allowing physicians to see, manage and analyse for further use the data collected from all tracked devices.

into the future. This allows patients to step in and take action before the peaks actually occur, and in this way avoiding the devastating effects that blood glucose extreme levels can cause to the body on the long run.

The decision management module is crucial to the requirements of this project, as most patients, even if they had a prediction for the BG levels for the next two hours, they most likely would not know what to do with it.

The project researchers have been careful to set very high safety criteria when it comes to the prediction rules and the advisory component. In the preliminary clinical trials, the predictive power of the system was shown to be useful up to two hours ahead. Its recommendations perfectly matched those of real doctors 88 % of the time and out of almost 1500 recommendations during testing it never provided any harmful advice.

For more information please visit www.diadvisor.eu

clinician application preview

DIAdvisor patient device

The software on the handheld device is self-learning. At first use, the prediction cone is less accurate.


Programming Language: C#

Microsoft Visual Studio on .NET Framework

Oracle Server

DIAdvisor patient device

As the input data history gets bigger, the prediction cone becomes utterly precise.

Advice Example

When the prediction curve is about to enter a dangerous area, a sound signal is received and the corrective advice is displayed on the monitor.
For more information please visit www.diadvisor.eu