Buchumschlag

Non-Invasive Glucose Level Monitoring from PPG using a Hybrid CNN-GRU Deep Learning Network

Gespeichert in:
Veröffentlicht in:arXiv.org (Nov 17, 2024), p. n/a
1. Verfasser: Soliman, Abdelrhman Y
Weitere Verfasser: Nor, Ahmed M, Fratu, Octavian, Halunga, Simona, Omer, Osama A, Mubark, Ahmed S
Veröffentlicht:
Cornell University Library, arXiv.org
Schlagworte:
Mean square errors
Error analysis
Deep learning
Monitoring
Machine learning
Root-mean-square errors
Artificial neural networks
Error detection
Glucose
Online-Zugang:Citation/Abstract
Full text outside of ProQuest
Tags: Tag hinzufügen
Keine Tags, Fügen Sie das erste Tag hinzu!
Abstract:Every year, humanity loses about 1.5 million persons due to diabetic disease. Therefore continuous monitoring of diabetes is highly needed, but the conventional approach, i.e., fingertip pricking, causes mental and physical pain to the patient. This work introduces painless and cheaper non-invasive blood glucose level monitoring, Exploiting the advancement and huge progress in deep learning to develop a hybrid convolution neural network (CNN) - gate recurrent unit (GRU) network to hit the targeted system, The proposed system deploys CNN for extracting spatial patterns in the photoplethysmogram (PPG) signal and GRU is used for detecting the temporal patterns. The performance of the proposed system achieves a Mean Absolute Error (MAE) of 2.96 mg/dL, a mean square error (MSE) of 15.53 mg/dL, a root mean square Error (RMSE) of 3.94 mg/dL, and a coefficient of determination (\(R^2\) score) of 0.97 on the test dataset. According to the Clarke Error Grid analysis, 100% of points fall within the clinically acceptable zone (Class A)
ISSN:2331-8422
Quelle:Engineering Database