Preemies, infants who are born too soon, have a higher incidence of Life-Threatening Events (LTE’s) such as apnea (cessation of breathing), bradycardia (slowing of heart rate) and hypoxemia (oxygen desaturation) also termed as ABD (Apnea, Bradycardia, and Desaturation) events. Clinicians at Neonatal Intensive Care Units (NICU) are facing the demanding task of assessing the risk of infants based on their physiological signals. The aim of this thesis is to develop a risk stratification algorithm using a machine-learning framework with the features related to pathological fluctuations derived from point process model that will be embedded into the current physiological recording system to assess the risk of life-threatening events well in advance of occurrence in individual infants in the NICU. We initially propose a point process algorithm of heart rate dynamics for risk stratification of preterm infants. Based on this analysis, point process indices were tested to determine whether they were useful as precursors for life-threatening events. Finally, a machine-learning framework using point process indices as precursors were designed and tested to classify the risk of preterm infants. This work helps to predict the number of bradycardia events, N, in the subsequent hours measuring point process indices for the current hour. The model proposed uses Quadratic Support Vector Machine (QSVM), a machine learning classifier, which can solve class optimization problems and execute data at an exponential speed with higher accuracy for risk assessment that might facilitate effective management and treatment for preterm infants in NICU. The findings are relevant to risk assessment by analyzing the fluctuations in physiological signals that can act as precursors for the future life-threatening events.
Date of publication
Dr. Premananda Indic, Dr. Mukul Shirvaikar and Dr. Jimi Francis
Master of Science in Electrical Engineering
Amperayani, Venkata Naga Sai Apurupa, "ASSESSMENT OF RISK IN PRETERM INFANTS USING POINT PROCESS AND MACHINE LEARNING APPROACHES" (2018). Electrical Engineering Theses. Paper 36.
Electrical and Computer Engineering Commons, Other Analytical, Diagnostic and Therapeutic Techniques and Equipment Commons, Other Biomedical Engineering and Bioengineering Commons, Risk Analysis Commons