[CCoE Notice] Thesis Announcement: Onyinyechi Ihesiulo, "Digital Twin of Glucose Metabolism of Type 1 Diabetes Patients"

[Greenwell, Stephen J]

[Thesis Defense Announcement at the Cullen College of Engineering]
Digital Twin of Glucose Metabolism of Type 1 Diabetes Patients
Onyinyechi Ihesiulo
April 25, 2025, 10 a.m. to 11 a.m. (CST)
Location: Engineering Bldg. 1, Small Conference Room or Teams Link<https://urldefense.com/v3/__https://teams.microsoft.com/l/meetup-join/19*3ameeting_MjdhOGY5ZTMtYTVlMS00MmI4LTkwNzQtY2NlMzUzN2RlOWUx*40thread.v2/0?context=*7b*22Tid*22*3a*22170bbabd-a2f0-4c90-ad4b-0e8f0f0c4259*22*2c*22Oid*22*3a*22f23897f8-9c83-4b3a-9670-72db9e1a2487*22*7d__;JSUlJSUlJSUlJSUlJSUl!!LkSTlj0I!C-dZQzfMMXj7PgkgDoqw0Ie5g9xui2Kf9dCHO6DW7MhVyq3ey4kxXywOva3J5GySdCdPUcz3an-zfwlpxK_iZMWUJN0$ >
Meeting ID: 288 044 602 529 4
Passcode: xr64gz3X

Committee Chair:
Marzia Cescon, Ph.D.
Committee Members:
Luca Pollonini, Ph.D. | Sandeep Gupta, Ph.D
Abstract
Digital twins (DTs) present an exciting potential in personalized therapy through the provision of in silico simulations of a patient's responses to treatments. This study introduces a novel blood glucose prediction and simulation framework that has been tailored to simulate the dynamics of individual glucose-insulin responses in people with Type 1 Diabetes (T1D) which can be exploited for the design, testing and regulatory approval of robust, precise, safe and effective novel insulin therapeutic regimes without the need ofany interventions on the individuals.

Our framework hinges on a model of glucose dynamics and a method for physiological parameter estimation which utilizes a Liquid Time-Constant Neural Network (LTCNN). The developed DT was trained and validated on synthetic data obtained from an established metabolic model of T1D dynamics, which provided a controlled environment consisting of 10 virtual subjects for evaluating the DT's prediction performance. We compared the LTCNN with Markov Chain Monte Carlo (MCMC) for the estimation of unknown parameters in the model, exploiting a method recently proposed in the literature.

 The MCMC-based DT prioritizes interpretable, physiology-driven parameter identification, whereas the LTCNN-based variation emphasizes data-driven dynamics modeling with real-time responsiveness. Both techniques were tested on five predefined in silico scenarios, and accuracy was measured using metrics such as MARD and RMSE.

According to the results, our DT is capable of accurately capturing the glucose behavior of an individual in the presence of variable insulin and meal inputs. The MCMC variation gives mechanical interpretability, whereas the LTCNN model provides fast inference and adaptable learning. This study contributes to the larger field of DT technology in diabetes management by highlighting the trade-offs between statistical inference and neuronal modeling in physiological systems.
[Engineered For What's Next]



-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://Bug.EGR.UH.EDU/pipermail/engi-dist/attachments/20250414/e71bfd00/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image001.png
Type: image/png
Size: 28468 bytes
Desc: image001.png
Url : http://Bug.EGR.UH.EDU/pipermail/engi-dist/attachments/20250414/e71bfd00/attachment-0002.png 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image002.png
Type: image/png
Size: 5699 bytes
Desc: image002.png
Url : http://Bug.EGR.UH.EDU/pipermail/engi-dist/attachments/20250414/e71bfd00/attachment-0003.png