Use this URL to cite or link to this record in EThOS:
Title: Quantifying brain maturation in the preterm baby from EEG sleep analyses
Author: Pillay, Kirubin
ISNI:       0000 0004 7966 1417
Awarding Body: University of Oxford
Current Institution: University of Oxford
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
The preterm (premature) baby, born before term-age of 37 weeks Postmenstrual Age (PMA, the age since the last menstrual cycle of the mother) remains vulnerable and is admitted to the Neonatal Intensive Care Unit (NICU). Continual improvement of care in the NICU has limited their lifelong impact. However, clinical interventions and other environmental stresses can facilitate delays in brain maturation resulting in abnormal neurodevelopmental outcomes (mental and motor disabilities) by school-age. This is not easily detected by existing methods in the NICU, though Electroencephalography (EEG) is emerging as a promising modality to capture such deviations through the maturation of sleep. Due to a shortage of trained clinicians and the infeasibility of continually labelling EEG, developing fully automated methods is crucial. Current methods are either not fully automated, do not consider the maturational effects across multiple sleep states, or have limited to no clinical validation on abnormal outcome data. Consequently, the aim of this research is to develop fully automated methods to quantify brain maturation from EEG sleep, through (i) robust automated sleep staging, (ii) tracking sleep-dependent developmental trajectories over PMA, (iii) assessing the quality of sleep-wake cycling with PMA, and (iv) validating these methods using both normal and abnormal outcome data. This thesis first presents a robust algorithm for detecting the Quiet Sleep (QS) and non-QS sleep states from EEG over a wide PMA range, with median sensitivity and specificity of 0.82 and 0.92, respectively. With this method, sleep states are then estimated from abnormal outcome data and a data-driven feature extraction applied to identify maturational 'biomarkers' and derive sleep-dependent trajectories over PMA. Developmental deviations are detected that identify abnormal outcome patients early. Probabilistic models are next developed to assess sleep-wake cycling by performing the first multi-state classification of sleep at term-age (with median sensitivity and specificity of 0.72 and 0.91, respectively). At this age, EEG differentiates into more complex states and the timing of this differentiation is also an important indicator of brain maturation. Finally, these models are expanded to novel, end-to- end solutions (employing Bayesian non-parametrics) that incorporate both age and sleep dependencies. These methods additionally classify severely delayed sleep-wake cycling behaviours, known as 'dysmaturity'.
Supervisor: Vos, Maarten De Sponsor: RCUK Digital Economy Programme
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Biomedical engineering ; Biomedical signal processing ; Machine learning