Date Thesis Awarded
Honors Thesis -- Access Restricted On-Campus Only
Bachelors of Science (BS)
Because of their underdeveloped immune systems, premature babies are at an increased risk to contract many illnesses. Thus, early detection of a disease is vital to saving a premature baby's life. Current methods of detecting illnesses, however, have been inadequate, providing many false positives and insufficient amount of warning time. However, patterns in the heart rate of babies have shown signs of predicting the onset of sepsis in premature infants. Research conducted by Prof. John Delos and others suggest that low variability and clusters of decelerations in an infant's heart rate may indicate an impending septic event. Additionally, there is weak evidence that low variability may be linked to gram-positive bacteria and clusters of decelerations may be linked to gram-negative bacteria. If this statement is true, then not only will the heart rate of an infant predict the onset of sepsis, but also provide a partial diagnosis and thereby indicate the preferred treatment for the baby. However, much more work needs to be done to prove this hypothesis. Over twelve terabytes of data has been collected on premature babies' heart rate and breathing. To search through this data, one first needs to know what to look for. Unfortunately, only looking for low variability and clusters of decelerations would be inadequate since most babies experience some low variability and decelerations in their heart rate at some point. Therefore, sophisticated statistical analysis is necessary to quantify this data. The general idea of this analysis includes creating many different heart rate characteristics (HRCs) and measuring their predictive power through multiple methods. The results of our research indicate that the HRCs of variance, sample entropy, and asymmetry are strong predictors of illness. However, no HRC shows strong signs of indicating the type of invading organism that caused the illness.
Dienstman, Evan, "Saving Babies Using Big Data" (2017). Undergraduate Honors Theses. William & Mary. Paper 1048.
On-Campus Access Only