Abstract
According to the statistics, millions of infants all around the world die because of bradycardia. Critical health conditions like this can be predicted and prevent using medical sensors, actuators, and artificial intelligence. Infants born before 37 weeks are considered premature and it occurs at a rate of 10% worldwide. Only in the USA, more than 1 in 10 pregnancies will end in preterm birth with critical heart conditions (CHD). About 25% of the preterm babies are affected by critical CHDs that about 40000 births per year. The general definition of Apnea is a pause in the regular breathing of baby lasting longer than 15-20 seconds. Normal breathing will vary but does not stop for any length of time. During sleep apnea, the infant's heart rate will go low and that's defined as Bradycardia. In this work, we focus on some specific events in cardiac and cardiorespiratory signal processing in real-time for preterm infants. Experimental results show that there is important information on heart rate variability (HRV) which can be extracted in real-time by the proposed algorithm and used for predicting bradycardia. The system also contains an actuator to stimulate the preemies to breathe, preventing sudden cardiac death. In addition, it can reduce the visiting frequency of a patient, and minor injuries/illness can be operated from a remote location. Therefore, we have developed a method for automatic detection of bradycardia and sleep apnea using heart rate, skin temperature and tri-axis acceleration data generated from non-contact biosensor. The proposed system can be used for home and hospital use. We performed supervised learning and extracted 23 features using time and frequency domain analysis to recognize pre- and post-life threating health conditions in patients. Feature selection procedures are used to reduce the number of features and processing time. For supervised learning, we compared three classifiers and selected the one with highest accuracy and sensitivity: decision tree, k-nearest neighbors (KNN) and eXtreme Gradient Boosting utilizing modified features. The results show that the proposed method can detect life threatening event for preterm infants in real-time with 99% accuracy. The numerical analysis is completed on data collected from 30 participants over a span of 4 months.