Sleep Apnea: Detection and Classification of Infant Patterns

Abstract Apnea is defined as a period in which an infant stops breathing. When the apnea period is greater than a critical period, serious damage in the infant's brain may result from lack of oxygen. In order to prevent this serious event, a need exists to find a reliable way to detect apnea and to determine causes of apnea. For that reason, sensors exclusively designed for research purposes, and especially manufactured apnea monitors with built‐in sensor, used at homes and pediatrics intensive care units in hospitals have been employed. As a result of the limitations and weaknesses of commercially available apnea monitors, in recent years, artificial intelligence techniques, especially neural networks have gained attention in terms of finding a reliable and alternative way to analyze respiration signals to determine if apnea exists and, if so, the length of apnea. In this study, an overview of apnea, difficulties in apnea monitoring, and elementary information for data acquisition and neural networks are given; a multi‐layered neural network for apnea recognition is designed; and applicability of the designed neural network trained with a set of respiration patterns obtained from infants who have apneic episodes is discussed in terms of detecting and classifying apnea patterns.