Simon is a DPhil (PhD) student in the paediatric neuroimaging group of Oxford University, where he investigates the inference of pain and physiological instability using electrophysiology in newborn infants. His research is jointly funded by the EPSRC and Reckitt Benckiser. He is interested in the broad applications of machine learning to clinical care.
Previously he worked as a clinical scientist in the NHS, both in Leeds Teaching Hospitals and the NHS Orthotic Research & Locomotor Assessment Unit. As a clinical scientist he provided combined technical and clinical support to neurological and orthopaedic rehabilitation units, and designed new medical devices, software and clinical systems. He completed an integrated Masters (MEng) in Biomedical Engineering at the University of Sheffield and a Masters (MSc) in Clinical Science at King's College London. He is a Chartered Engineer and a registered Clinical Scientist in the UK.
Mechanical ventilation by endotracheal intubation is often required in preterm infants. Too little ventilation can result in respiratory distress and mortality. Too much ventilation also has risks as acute lung injury can be caused both by the mechanical forces of ventilation, and by the inflammation which follows. Balancing these risks is important, however it is clinically difficult to determine the optimum time to extubate (i.e. remove intubation from) an individual infant, with around 25% of extubated preterm infants requiring re-intubation within a short time-frame. This re-intubation is sometimes called "extubation failure", and can be distressing and harmful for infants. Predicting when extubation will fail could avoid this.
Simon is using time-series physiological data from newborn infants undergoing mechanical ventilation by endotracheal intubation, together with information from hospital notes on infant outcomes. This data will come from the MIMIC-III Waveforms database, a large dataset of physiological data and matched hospital notes data from patients admitted to ICU. This project will use a simple machine-learning model to classify infants according to likelihood of being in the group of infants who later suffer from extubation failure. This will provide a predictor of extubation failure from the vital signs of mechanically ventilated, preterm neonatal infants.