Dual smart sensor data-based deep learning network for premature infant hypoglycemia detection

In general, deficient birth weight neonates suffer from hypoglycemia, and this can be quite disadvantageous. Like oxygen, glucose is a building block of life and constitutes the significant share of energy produced by the fetus and the neonate during gestation. The fetus receives glucose from the placenta continuously during gestation, but this substrate delivery changes abruptly, and the fetus's metabolism changes significantly at birth. Hypoglycemia is one of the most frequent pathologies affecting the change of newborns in neonatal critical care units. This work is now introducing a system, HAPI-BELT, empowered by dual intelligent sensors and Deep Learning (DL) algorithms for tracking and continuously detecting hypoglycemia in preterm newborns. This article comprises a smart belt with an intelligent camera and photoplethysmography (PPG) attached. This device tracks changes in the infant's motion, skin colour, and breathing patterns; this is done through a PPG sensor strapped either on the belly or chest of an infant, logging information on heart functioning. The digital data gathered by this PPG sensor and image data captured from the smart camera are then processed by a Raspberry Pi Zero 2 W. It does most of the data analysis and decision-making. Feature Extraction (FE) is done through CAT-Swarm Optimization. Based on features, the sorted-out data gets evaluated through a GRU-LSTM (Gated Recurrent Unit - Long Short-Term Memory) network to identify the state of the infant as usual and suggestive of hypoglycemia-blood glucose below 70 mg/dL, pale complexion, profuse perspiration. When hypoglycemia is identified, an alert is sent to the medical professionals to take necessary action with utmost urgency. Therefore, an integrated approach ensuring timely medical interventions and real-time monitoring can help better outcomes for preterm newborns.