-
1رسالة جامعية
المؤلفون: Weller, Jennine Hess Dearolf
المساهمون: Caffo, Brian, Haut, Elliott R, Hackam, David J, Bembea, Melania M, King, Betsy A
مصطلحات موضوعية: necrotizing enterocolitis, prediction, machine learning, surgery
الوصف: Necrotizing enterocolitis (NEC), a devastating disease of premature bowel, is challenging to predict. The disease is rare, with incompletely understood pathogenesis, rapid onset and progression, and insufficient diagnostic criteria. Using a systematic review of the literature, a cultivated dataset of published neonatal radiographs, and a publicly available neonatal critical care database, this dissertation examines novel approaches to improve predictions of NEC. First, in a review piece, we summarize surgical care for patients with NEC (Chapter 2). We provide a foundational framework to understanding NEC by describing the diverse presentations of the disease and discussing current best practices to reduce NEC-associated morbidity and mortality. Second, we conduct a systematic review of published prognostic models for predicting NEC onset and progression in hospitalized infants (Chapter 3). We find that published models have fair to poor discrimination of NEC outcomes and high risk of bias, limiting model clinical utility. Third, we develop an image classifier to support surgical resident recognition of pneumatosis intestinalis, a radiographic sign of NEC (Chapter 4). We find that a deep convolutional neural network trained on neonatal abdominal radiographs can successfully detect pneumatosis and performs comparably well to senior surgical residents. Fourth, we use the MIMIC III Clinical Database to develop an early warning score for NEC based on routinely available clinical data during an infant's stay in a neonatal intensive care unit (NICU) (Chapter 5). We find that models accurately predict NEC before disease onset, with first NEC risk detection occurring days previously. Fifth, in a perspective piece, we reflect on the promises and challenges of utilizing machine learning methods in NEC prediction and research (Chapter 6). We advocate for policy and practice changes to improve NEC prediction efforts. Overall, this dissertation highlights strengths and limitations of existing NEC prediction models and offers ...
وصف الملف: application/pdf
-
2رسالة جامعية
المؤلفون: Bose, Sanjukta Nandi
المساهمون: Winslow, Raimond L, Sarma, Sridevi V, Bembea, Melania M, Mallada, Enrique
مصطلحات موضوعية: prediction, precision medicine, machine learning, early warning, dynamical system, model predictive control, feedback control, ICU, multiple organ dysfunction, organ dysfunction, MODS, risk score, cardiac arrest, congenital heart disease, respiratory failure, mechanical ventilation, COVID-19, ARDS, oxygen saturation, SpO2, FiO2, PEEP, ventilator settings
الوصف: Adverse clinical events like cardiopulmonary arrest and multiple organ dysfunction are life-threatening events with potentially debilitating health outcomes and are a leading cause for mortality in intensive care unit (ICU) patients. Traditionally, clinical decision making is based on periodic review of physiologic monitoring data by attending clinical staff. Modern hospitals collect an enormous amount of information every minute, but making inferences consistently on a continuous-time basis is impractical for clinicians. Instead, data-driven prediction models can be employed to extract value from this continuous stream of information and provide early warning of adverse events to facilitate timely therapeutic intervention and allow efficient patient monitoring. For patients already in critical condition, patient-specific dynamical system models with feedback control can suggest optimal intervention strategies. This dissertation has four contributions. The first three are the development and validation of prediction models for early recognition of increased risk of (a) cardiac arrest in infants with congenital heart disease, (b) pediatric multiple organ failure, and (c) need for invasive mechanical ventilation due to respiratory failure in pediatric ICU patients using machine learning methods. In all three, we observed that the time-varying risk in individual patients increased several hours before an observed event with a median early warning time of 17, 23, 10 hours respectively. We then investigated the presence of groupings within the positive predictions and found 2-3 distinct clusters where the high-risk group had >90% precision. Additionally, we also demonstrated a novel approach for combining medication history-based features to improve prediction performance. The fourth contribution was the application of system identification to learn the dynamics of oxygen saturation in COVID-19 patients with acute respiratory distress syndrome (ARDS) and implement optimal control strategies for maintaining desired ...
وصف الملف: application/pdf
-
3رسالة جامعية
المؤلفون: Chyn, Michelle
المساهمون: Sarma, Sridevi V, Winslow, Raimond L, Bembea, Melania M
مصطلحات موضوعية: multiple organ dysfunction syndrome, time series data, early detection, generalized linear model, pediatric
الوصف: Multiple organ dysfunction syndrome (MODS) has an incidence rate of between 11 to 56\% in the PICU. Early prevention and treatment of MODS is important in the pediatric population as it increases mortality and leads to possible negative functional outcomes in adulthood. MODS severity is measured using a few different metrics, among which the Pediatric Logistic Organ Dysfunction 2 Score (PELOD-2) is the most recent, pediatric multi-center validated scoring system. This study attempted to build a generalized linear model to detect risk of PICU patients at Johns Hopkins Children's Center from a retrospectively gathered cohort, using PELOD-2 Score>=6 to define MODS severity and minute to minute physiological data as model covariates. Patient specific models were built with a two hour window for transitioning into severe state, the positive class, and the non-severe state was undersampled to balance classes. A global model was built across the majority of the patient population with similar parameters in order to create a more useful, clinical applicable model. The accuracy, sensitivity, and specificity of training and testing sets were calculated for each model. Patient specific models performed well, but performance decayed for the global model, where predictions at the patient level for risk of transitioning had high sensitivity and very low specificity. Future research should continue to refine the definition of a severe state of MODS and calibrate the sampling scheme with regards to ratio of data points labeled as healthy versus at risk in order to improve global model performance.
وصف الملف: application/pdf
