Predictive modeling of hypoglycemic events in type i diabetes through artificial neural networks
Predictive modeling of hypoglycemic events in type i diabetes through artificial neural networks

Author(s): Usic Ghenadie
Subject(s): Social Sciences, Sociology, Health and medicine and law
Published by: Biblioteca Ştiinţifică a Universităţii de Stat Alecu Russo
Keywords: artificial neural networks; predictive modeling; hypoglycemia; generic algorithms; glucose monitoring
Summary/Abstract: This paper presents an approach for predicting hypoglycemic events in patients with Type I diabetes using artificial neural networks (ANNs). Hypoglycemia represents a significant complication often induced by intensive insulin therapy, with nocturnal episodes being particularly perilous due to the obscuring effect of sleep on early symptoms, which may escalate to seizures, coma, or even death. The study analyzes 3 months of continuous glucose monitoring (CGM) data, insulin dosage logs, and meal records from five individuals diagnosed with Type I diabetes. By employing historical data, the propored methodology trains a deep learning model to identify patterns that preemptively signal the risk of hypoglycemia. Furthermore, the paper shows a way to enhance the sensitivity and specificity of predictions by integrating genetic algorithms into the training process. This adaptation not only refines the model’s predictive accuracy but also enables it to dynamically adjust to the unique physiological profiles of individual patients, thereby enhancing the timeliness and reliability of its predictions.

• Details
• Contents