KalmanNet application in Continuous Glucose Monitor Estimation

Huang, Ash (2025) KalmanNet application in Continuous Glucose Monitor Estimation. Bachelor's Thesis, Artificial Intelligence.

Preview	Text AshHuangBachelorThesis.pdf Download (2MB) | Preview
	Text Toestemming.pdf Restricted to Registered users only Download (177kB)

Abstract

Continuous glucose monitoring (CGM) is critical for diabetes management, yet translating interstitial glucose (IG) measurements into accurate blood glucose (BG) estimates remains challenging due to physiological lag and signal noise. While state-of-the-art methods, such as Particle Filters (PF), can be accurate, their computational intensity and dependency on high accuracy models can limit their utility for application. This thesis explores the application of KalmanNet (KN), a deep learning-based Kalman filter, as an alternative solution. We adapted the KN architecture and trained it on simulated clinical CGM data to assess its performance profile. Our evaluation suggests that while KN achieves accuracy comparable to a PF, its primary advantages lie in its efficiency and robustness. KN’s higher robustness and a drastic reduction in inference time establish KalmanNet as a potential practical solution for the CGM task.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Cardenas Cartagena, J. D.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	09 Sep 2025 07:34
Last Modified:	09 Sep 2025 07:34
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/37007

Actions (login required)

View Item