Wed 28th November, 2012
Medical experts could soon be able to further identify newborn babies at risk of brain development issues using technology developed at the University of Bedfordshire.
A computerised system designed by Livija Jakaite to accurately estimate patterns of the brain from electroencephalogram (EEG) tests (brain readings) could help doctors "save the lives" and prevent damage in later life.
More than 70,000 babies per year are born premature or sick and their brain development can suffer from birth injuries.
Currently ultrasound scanning is the best way of diagnosing these injuries, however newborn babies may still be at risk as the brain develops and this can only be recognised in EEGs.
Ms Jakaite, who did her PhD in Science and Technology at the University's Institute for Research in Applicable Computing, said: "Age-related patterns are difficult to recognise as they widely vary during first weeks after birth, and newborn EEG is a very weak signal easily corrupted by artefacts and electrode noise.
"This makes the analysis of EEG laborious, and experts spend hours to confidently interpret an EEG recording. As a consequence of that, EEG analysis cannot be made available for all newborns at risk.
"An expert assessment can sadly be made mistaken, as there are no standardised rules for interpretation of brain maturity patterns and for estimating the risk of decision making."
To deal with the problem of interpretation, various methods for computer-assisted assessment of maturity-related patterns were developed. Using methods of probabilistic inference is crucial for experts to quantitatively evaluate the confidence of assessments. That can be estimated most accurately by using the Bayesian theory of probabilistic inference. This technology, however, requires massive computations to explore assessment models, and has only recently been made computationally feasible.
Ms Jakaite said: "Using the Bayesian theory, we developed a new technology for EEG assessment of newborn brain development and tested it on almost 1000 healthy newborns aged between 36 and 45 weeks post-conception.
"We found that the accuracy of the proposed technology is comparable with that provided by the most experienced experts. Overall the proposed technology provided the highest accuracy of EEG assessment along with accurate estimates of risk of decision making.
Analysing the risk, experts can decide how confident they are in making decisions."
The technology also provides experts with information on how decisions are made. The proposed technology enables experts to timely identify newborns at risk of brain development pathologies and thus save their health and lives.
The £80,000 work – funded by The Leverhulme Trust with EEG date from the University of Jena, Germany - was deemed so impressive that Ms Jakaite, 27, was shortlisted for an engineering award at the National Science + Engineering Competition – becoming the University's first nominee since Professor Carsten Maple, currently Pro Vice Chancellor for Research and Enterprise.
Ms Jakaite, who also has a BSE in Artificial Intelligence, was also invited to present her finding in the House of Commons.