|M.Sc Student||Azriel Raphael|
|Subject||Machine Learning to Support Triage of Children with|
Epileptic Seizures in the Pediatric Intensive
|Department||Department of Biomedical Engineering||Supervisors||DR. Joachim Behar|
|DR. Danny Eytan|
|Full Thesis text|
Introduction: Epileptic seizures are relatively common in critically-ill children admitted to the pediatric intensive care unit (PICU) and thus serve as an important target for identification and treatment. Most of these seizures have no discernible clinical manifestation but still have a significant impact on morbidity and mortality. Children that are deemed at risk for seizures within the PICU are monitored using continuous-electroencephalogram (cEEG). cEEG monitoring cost is considerable and as the number of available machines is always limited, clinicians need to resort to triaging patients according to perceived risk in order to allocate resources. This research aims to develop a computer aided tool to improve seizures risk assessment in critically-ill children, using an ubiquitously recorded signal in the PICU, namely the electrocardiogram (ECG). Previous works included the development of algorithms for individual seizure detection. This resulted in high false alarm rates (range of 1.1-9.5/h) making the approach clinically impracticable. We propose a novel patient-level approach. Methods: We hypothesize that the effect of seizures in comatose children will be reflected by specific patterns on their ECG at the time of the seizures and beyond. The database includes ECG recording of 176 children with a reduced level of consciousness. Among these patients, 52 children experienced seizures. The patients were divided into train and test set with a ratio of 2/3 and 1/3 respectively. Heart-Rate Variability (HRV) and ECG morphological (MOR) features were engineered from the ECG signal. A data-driven machine learning model was developed, based on the HRV and MOR features extracted from the first hour of ECG recording and the clinical data of the patient. Results The most predictive features were the age of the patient, the brain injury as coma etiology and the QRS area. For patients without any prior clinical data, using one hour of ECG recording, the classification performance of the random forest classifier reached an area under the receiver operating characteristic curve (AUROC) score of 0.84. When combining ECG features with the patient’s clinical history, the AUROC reached 0.87. Perspective: Taking a real clinical scenario, we estimated that our clinical decision support triage tool can improve the positive predictive value by more than 59\% over the clinical standard.