Automated assessment of cerebral autoregulation and coupled dynamics using NIRS & EEG scoring
The goal of the PhD project is to develop software for automated EEG scoring, cerebral autoregulation (CAR) assessment and quantification of linked dynamics between EEG activity and cerebral oxygenation (NIRS), for use in computer-aided prediction of neurological outcome in neonates in neonatal intensive care units (NICUs). More specifically, the PhD student will develop advanced algorithms (graphs using appropriate kernels, multiscale entropy, tensor-based Blind Source Separation, fractals, subspace projections) for the joint analysis of EEG and NIRS measurements. Time-frequency features will be derived as a marker for brain recovery (after any damage), and will also be used for the automatic detection of sleep stages in newborns. NIRS measurements (surrogates for cerebral blood flow) are used to assess cerebral autoregulation (CAR) via advanced nonlinear methodologies. The combination of all extracted features is used as input to clinically interpretable machine learning classifiers (Support Vector Machines, convolutional neural networks) to determine the severity of EEG abnormalities, and evaluate the combined use of EEG and NIRS scores as biomarkers for (non)pathological condition. These scores are used also to develop functional growth charts for normal neural development in neonates/infants. The PhD student will develop the software toolbox to quantify the linked dynamics between EEG and NIRS measurements which should run in real-time at bedside in NICUs and be tested on clinical data provided by hospitals in Helsinki and Utrecht.