Back to search

NAERINGSPH-Nærings-phd

Modellering for sanntids estimering og identifisering av glukose metabolismeprosesser i mennesket.

Alternative title: Modelling, real-time estimation and identification of human glucose metabolism.

Awarded: NOK 1.5 mill.

Project Number:

242167

Project Period:

2015 - 2018

Funding received from:

Location:

Diabetes is a disease that grows epidemically. More than 415 million people are estimated to have the disease by now. The effect of diabetes is strongly deteriorated blood sugar (glucose) control with death as the ultimate outcome. Poor blood glucose control results in other diseases such as nerve damage, cardiovascular diseases and blindness. To control the disease, insulin has to be injected. To inject the proper amounts, the blood glucose level has to be measured regularly, usually by methods where the skin is penetrated for direct blood access (invasive methods). Prediktor Medical develops a glucose measuring device for non-invasive and continuous glucose measurement. The sensor is a wearable device to be worn on the arm. The operating principle is to combine a set of measuring principles, such as near infrared light transmission, body/arm acceleration and electrical properties of tissue, and combine these measurements with predicted values from a mathematical model running in real-time inside the device. The simulator reproduces the metabolism of blood glucose, and acts as a synthetic sensor. The synthetic sensor values are combined with the other physically based measurements to improve the glucose estimate. The Industrial PhD has focused on the development and refinement as well as on analysis of such mathematical models of the body metabolism to act as a synthetic blood glucose sensor. This model will simulate appearance, utilization, storage and generation of glucose in the body. A challenge is to make the model simple and robust for such real-time use. The interplay between the model and the physical measurements is part of the PhD work as well. Models of glucose dynamics are often used in very controlled settings like clinical research experiments, where everything the test subject does, like eating or injecting insulin, is accurately logged. In free-living settings, a quite different situation is the case; the user lives a normal life and can not be expected to log accurate and complete data at all times. This makes it necessary to be able to estimate and correct missing information in near real time based on measurements. A method to estimate missing meal information based on data from continuous glucose monitors (CGM) was developed and tested as part of the PhD project. In addition, a method to automatically smooth and detect errors in glucose data was developed, tested and published (https://doi.org/10.1109/JBHI.2018.2811706). It is important to adapt the model complexity to the intended use. A complex model intended for simulation use is in many cases unsuitable for use in a real-time estimation setting. The information content in data from diabetes-related equipment like CGM and insulin pumps is also often lower than what is collected in clinical research studies. This makes it important to determine which parameters of a model that are identifiable from a given set of data. This PhD project has also investigated and tested methods for identifiability analysis in glucose dynamics models. The characteristics of different types of glucose measurements is also an important aspect, which must be determined to use the measurements, both for the users of the system and their caretakers, and for researchers intending to use the measurements, either for model identification or as reference measurements to test other systems against. E.g. it is commonly the case that measurements from CGMs have a bias and a lag compared to fingerprick blood glucose measurements. A study of Freestyle Libre from Abbott, a commercially available system for blood glucose monitoring, was performed and published as part of the PhD project (https://doi.org/10.3390/bios8040093).

-

Prosjektet er sterkt knyttet til virksomheten i Prediktor Medical AS. Sistnevnte firma er et nystartet firma, hvis mål er å utvikle og kommersialisere ikke-invasive (ikke blodige) måleinnretninger for kontinuerlig måling an blodsukker hos mennesker. Prediktor Medical er finansiert via tilskudd fra Oslofjordfondet, Innovasjon Norge og private investorer. Prediktor Medical er nå i en fase for å finne en industriell eller en finansiell aktør for videre finansiering mot et kommersielt produkt. Produktet er en klokkelignende måler som bæres rundt håndleddet og som måleteknisk baseres på en kombinasjon av spektroskopi, bioimpedans, akselerometre og en matematisk modell som integrerer alle sensorsignalene ved hjelp av en metabolisme-modell av blodsukker- og energibalansen i kroppen. Dette produktet (heretter kalt GlucoPred) skal fokuseres mot to ulike markeder: 1) Det medisinske markedet for bruk av diabetespasienter. Måleren som skal utvikles vil revolusjonere dette markedet som er enormt (i størrelsesorden 80 milliarder NOK per år). 2) Helse- og fitnessmarkedet hvor energiomsetning, blodsukker-nivå, aktivitetsnivå, laktatnivå og vektutvikling er interessante parametre å overvåke kontinuerlig. Dette er også et meget betydelig marked på flere milliarder årlig. Teknologien som utvikles muliggjør ikke-blodige målinger av idrettsviktige parametre så som laktat. Dette kan ikke gjøres i dag uten å ta en blodprøve.

Publications from Cristin

No publications found

No publications found

Funding scheme:

NAERINGSPH-Nærings-phd