Classification of Electroencephalogram in Ovine Model of Perinatal Hypoxia by Movement of the Autoregressive Model Parameters
MetadataShow full item record
Hypoxic-ischemic encephalopathy (HIE) secondary to perinatal asphyxia occurs when the brain does not receive enough oxygen and blood. Oxygen deprivation can be caused by knotting of the umbilical cord, failure to breathe at birth, as well as a difficult birth. HIE is a dangerous condition that requires immediate medical intervention, e.g., therapeutic hypothermia. HIE due to acute perinatal asphyxia remains an important cause of mortality and neurodevelopmental impairment in childhood. HIE accounts for mortality numbers in excess of 840,000 annually or 23% of all neonatal deaths worldwide and much more significant rates of morbidity. Therefore, a surrogate marker for “intact survival” is necessary for the management of HIE. The challenge is early detection of the severity of HIE at the point-of-care. The severity of HIE can be graded based on the clinical presentation including the presence of seizures using a clinical classification scale called Sarnat staging. The clinician identifies infant’s alertness, muscle tone, seizures, pupils, respiration, and duration of the illness to determine Sarnat grade of mild, moderate, or severe. However, Sarnat staging is subjective, and the score changes over time. Also, seizures are difficult to detect clinically and are associated with poor prognosis. Therefore, a tool for continuous monitoring at the cot-side is necessary, e.g., conventional electroencephalogram (EEG) that non-invasively measures the electrical activity of the brain from the scalp. Full multi-channel EEG with at least nine and a maximum of 40 electrodes on the scalp and face takes at least 60 minutes of careful handling by trained personnel. Amplitude-integrated EEG (aEEG) is more straightforward with three (one differential channels) to five (two differential channels) electrodes attached to the scalp and highly correlates with conventional EEG in identifying abnormal power in the signal. Therefore, the aim of this project was to construct a prognostic algorithm based on EEG, to differentiate between normal, hypoxic, and ictal states in a perinatal ovine model of hypoxia with low computational expense applicable to limited resource setting. The algorithmic prognosis is done by classifying the parameters of a fourth order autoregressive (AR) model on a two electrode differential EEG measuring system, with varying states of oxygenation by tracking the position of the AR parameters relative to previously defined decision planes. In this project, I closely integrated computational modeling and EEG neuro-monitoring with animal studies, to investigate the global neuronal state during birth asphyxia and resuscitation towards clinical translation in the future. Our AR based algorithm can be easily implemented in a mobile handheld device that is under development for early classification of HIE severity at the point-of-care.