Type 1 diabetes (T1D) is an autoimmune disease that affects millions of people worldwide. A most challenging aspect regarding diabetes therapy is the way to calculate the insulin bolus amount to inject before meals. The artificial pancreas (AP), combining both blood glucose monitoring and automatic insulin delivery, has demonstrated its effectiveness in T1D treatment. However, one of the limitations of current AP devices is the fact that the patient needs to insert manually the amount of insulin to be released and the bolus is calculated on the estimated carbohydrate intake, while other nutritional factors of the patient's meal are not taken into account. To overcome this issue, in this paper, two innovative algorithms to predict the postprandial blood glucose concentration after the meal in T1D patients on AP systems are presented. The proposed algorithms, which cover a time span of prediction of 180 minutes, take into account not only the carbohydrates amount in the meal but also other selected nutritional factors. More specifically, the proposed algorithms are based on feed forward multi-layer neural networks (FFNNs) and long short-term memory networks (LSTMNs) with a specific hyper-parameter configuration. The output of the proposed architectures consists of a predicted glycemic curve. The algorithms were validated by comparing the predictions of the networks performed on a test dataset with the measured glycemic values recorded by the hybrid closed-loop systems worn by the patients.

Neural Network-Based Prediction and Monitoring of Blood Glucose Response to Nutritional Factors in Type-1 Diabetes

Andrea Cataldo;
2022

Abstract

Type 1 diabetes (T1D) is an autoimmune disease that affects millions of people worldwide. A most challenging aspect regarding diabetes therapy is the way to calculate the insulin bolus amount to inject before meals. The artificial pancreas (AP), combining both blood glucose monitoring and automatic insulin delivery, has demonstrated its effectiveness in T1D treatment. However, one of the limitations of current AP devices is the fact that the patient needs to insert manually the amount of insulin to be released and the bolus is calculated on the estimated carbohydrate intake, while other nutritional factors of the patient's meal are not taken into account. To overcome this issue, in this paper, two innovative algorithms to predict the postprandial blood glucose concentration after the meal in T1D patients on AP systems are presented. The proposed algorithms, which cover a time span of prediction of 180 minutes, take into account not only the carbohydrates amount in the meal but also other selected nutritional factors. More specifically, the proposed algorithms are based on feed forward multi-layer neural networks (FFNNs) and long short-term memory networks (LSTMNs) with a specific hyper-parameter configuration. The output of the proposed architectures consists of a predicted glycemic curve. The algorithms were validated by comparing the predictions of the networks performed on a test dataset with the measured glycemic values recorded by the hybrid closed-loop systems worn by the patients.
File in questo prodotto:
Non ci sono file associati a questo prodotto.

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11587/473904
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
social impact