Newborn Health: RCGAN Study

The aim of the PhD project is to improve training for medical professionals who monitor babies in the NICU through the synthesis of realistic neonatal seizure time series using recurrent conditional generative adversarial networks.

Thousands of hours of neonatal health recordings will be used to create a software programme and documentation that will employ AI to simulate realistic sensor data that can be integrated into training simulation tools used by medical staff. 

Newborns in the neonatal intensive care unit (NICU) are monitored with a range of sensing technologies  (video, EEG, ECG, respiration, BP) in an attempt to detect harmful neurological dysfunction when it occurs. Interpretation of the data produced requires extensive training on the part of the neurophysiologist as the signatures of each specific condition can be subtle. In developing countries, there is a severe shortage of the trained personnel required and to compound matters, there is a lack of instruction tools based on realistic data sources to enable the most effective use of their training. 

The purpose of this work then is to develop software that will use techniques from artificial intelligence to simulate realistic sensor data representative of the type seen in the NICU including presentation of seizure events with an appropriate range of variability. The nature of this work is computational modelling and we will not require the collection of any new data for this purpose. Instead, we will avail of the collaboration established between Prof Ward and Prof Geraldine Boylan and the Infant team at University College Cork who has collected hundreds and thousands of hours of the type of data required for this study and are one of the world’s leading research groups in neonatal health. 

In terms of techniques, we propose to use methods from machine learning called Recurrent Conditional Generative Adversarial Networks (RCGAN) to synthesize realistic multi-modal time series fully representative of healthy and pathological datasets as monitored in the NICU. These technologies will be trained through exposure to large, comprehensive datasets available within the collaboration while the final results will be tested both for fidelity of realism and protection of privacy of the generating sources. The final result will be a piece of open access software and documentation which can be integrated into third party training simulation tools.  


Sleeping baby

How To Apply

Candidates applying for this studentship must have a first or second class degree or Masters in a relevant discipline. Previous experience in a health-related discipline is expected and relevant experience is welcomed.

A stipend of 18,500 euro per annum and a cover for UCC registration fees will be available for the successful applicant (non-EU members may apply but additional fees are paid by the candidate).  The position is funded for 4 years with an allowance for travel and publication costs. 

The successful applicant will be asked to proceed through the Postgraduate Applications Centre (PAC, for formal registration.

Interested candidates should submit:

  1. a covering letter detailing their interest and reasons for applying for this position 
  2. (a CV (including the name and contact details of 2 referees)
  3. a summary of their relevant research experience to date

The above documents should be emailed to on or before the closing date of 11 July.

Please reference ‘PhD Studentship RCGAN Study’ on the subject line of your application.

The successful applicant is expected to commence the post by 1 October 2018. 

Further Enquiries: please contact Prof Tomás Ward

Subscribe to our newsletter