Abstract
Preterm birth is a significant global public health concern, affecting millions of babies yearly. Despite advancements in medical care that have improved the survival rates of preterm infants, preterm birth remains a leading cause of neonatal morbidity and mortality worldwide. It has both short-term and long-term health consequences that can profoundly impact the child's growth and development, as well as their family and society.
One of the challenges preterm infants face is their underdeveloped immune system, which makes them more vulnerable to infections and other health problems. Their delicate condition requires specialized care, often provided in a Neonatal Intensive Care Unit (NICU). In NICUs, physiological signals such as heart rate and blood oxygen levels, also known as SpO2 levels, are monitored to ensure the well-being of these infants. However, there is currently no standardized technique for quantitatively evaluating the progress and development of preterm infants in the NICU.
Hypoxemia is a significant concern among the health complications associated with preterm birth. Hypoxemia refers to low oxygen levels in the blood and is a common issue in preterm infants. It is linked to a range of short-term and long-term health problems, including retinopathy of prematurity, neurodevelopmental disorders, and even mortality. Evaluating the severity of Hypoxemia and measuring the effectiveness of treatment require using indices related to its frequency, patterns of occurrence over time, and the total time spent in a hypoxic state.
Hypoxemia in preterm infants is associated with various complications, including bronchopulmonary dysplasia (BPD). BPD is a chronic lung disease that primarily affects preterm infants who require mechanical ventilation or oxygen therapy. It typically develops due to lung injury caused by mechanical ventilation and inflammation in the developing lung. Infants with BPD have trouble breathing and may require ongoing respiratory support, such as oxygen therapy or mechanical ventilation. BPD can also lead to long-term health complications, including impaired neurodevelopment, pulmonary hypertension, and an increased risk of respiratory infections. Currently, no established techniques are considered optimal for screening infants with evolving, established, or severe BPD. Developing quantifying measures that can predict BPD earlier than four weeks after birth would be highly beneficial for clinical settings.
However, developing a predictive model that customizes treatment plans based on an infant's previous oxygenation behavior could assist clinicians in anticipating the infant's condition and tailoring interventions accordingly. It is crucial to understand the temporal processes underlying hypoxemic events and analyze data on their frequency and distribution of inter-event intervals.
The first objective of this work is to analyze hypoxemic events in preterm infants and gain insights into the temporal dynamics underlying these events. This analysis aims to derive statistical models describing the occurrence and frequency of Hypoxemia in preterm infants. By studying the probability distribution of the inter-hypoxemia interval (IHI), and the duration between two successive hypoxemic events, valuable insights into the system dynamics can be obtained.
The second objective is to develop a signal-based model by incorporating IHI distribution features obtained from the analysis of data collected during the first week of life. This model aims to identify infants likely to be diagnosed with BPD. Early detection of BPD would enable timely interventions and improved clinical decision-making, potentially leading to better outcomes for preterm infants.
In conclusion, the stochastic properties of time intervals between adverse hypoxemia events were investigated in preterm infants. Theoretical distributions with observed data fit in and found high accuracy in characterizing these intervals. The analysis revealed that different distributions were suitable for capturing dynamics over short and long-time scales. Inter-hypoxemic interval (IHI) distributions were explored as potential indicators for predicting BPD. Remarkably, using only IHI parameters, the prediction model achieved an impressive 91% sensitivity for identifying BPD subjects. The findings suggest that IHI parameters can serve as early precursors for BPD prediction, providing clinicians with valuable insights during the first week of an infant's life. Furthermore, the combination of IHI parameters with demographic factors like gestational age and birth weight yielded the best predictive model. These results contribute to our understanding of hypoxemia dynamics and offer valuable tools for managing the risks associated with BPD in preterm infants.
Date of publication
Fall 7-25-2023
Document Type
Thesis
Language
english
Persistent identifier
http://hdl.handle.net/10950/4247
Committee members
Premananda Indic, Ph.D, Prabha Sundaravadivel, Ph.D., Shawana Tabassum, Ph.D.
Degree
Masters in Electrical Engineering
Recommended Citation
Mukherjee, Ratri, "A MACHINE LEARNING APPROACH FOR THE EARLY DETECTION OF BRONCHOPULMONARY DYSPLASIA (BPD) IN PRETERM INFANTS USING INTER HYPOXEMIA INTERVALS" (2023). Electrical Engineering Theses. Paper 57.
http://hdl.handle.net/10950/4247