Preterm Enhanced Automated Capture of Comfort Knowledge (PEACOCK)
Research type
Research Study
Full title
Preterm Enhanced Automated detection and assessment of pain and comfort levels in preterm babies using video and computerised deep learning (Peacock study)
IRAS ID
299441
Contact name
Janet Berrington
Contact email
Sponsor organisation
The Newcastle upon Tyne Hospitals NHS Foundation Trust Newcastle Joint Research Office
ISRCTN Number
ISRCTN00000000
Duration of Study in the UK
1 years, 0 months, 31 days
Research summary
Research Summary:
Preterm infants may experience significant pain and discomfort as a necessary part of providing high quality medical care. We aim to conduct an observational pilot study exploring the use of computer machine learning methods to determine the comfort state of preterm infants in a hospital, in particular whether they may be experiencing pain/discomfort and resting/comfort states. The study will investigate which behavioural and physiological indicators are most informative of infant’s state. We will record video/audio footage of the baby's face and movements and relate these to any routine procedures the baby has such as blood tests, or comparing differences between when they are resting in an incubator compared to having skin to skin comfort care with their parents.
Traditionally, neonatal pain is assessed intermittently by bedside nurses using their experience, supplemented with recording of standardised 'pain scales' but these methods have high inter-observer variability. More accurate and valid determination of infant pain or discomfort using automated techniques may improve the baby's overall experience, and improve the rigor with which pain is determined, and therefore treated.
Video/audio footage will be analysed by a computer using so-called 'machine learning models' which can analyse large volumes of data to determine how for example small movements of the limbs, or subtle changes in facial expressions may relate to pain or discomfort.
There are no changes to the patient journey and no risks or discomfort to the baby. All normal medical and nursing procedures will be carried in the normal way. We will use a camera/microphone mounted on a tripod, or placed on top of an incubator to record limb and facial movements and sound, and also collect data from standard bedside monitoring. The study will be conducted on Ward 35, neonatal intensive care Unit, Royal Victoria Infirmary Newcastle Hospitals NHS Trust.
Summary of result:
This study aimed to develop a dataset suitable to be used in training an automated system for monitoring pain in preterm infants and to improve the accuracy of pain assessments. There are not yet any results, the main findings and the effectiveness of the machine learning models will be reported once the analysis is complete. The results will be published and disseminated as part of a PhD thesis and in relevant journals.Summary of Results
The PEACOCK trial was sponsored by Newcastle upon Tyne Hospitals NHS Foundation Trust and funded by the Leverhulme Trust as part of the Newcastle University doctoral training program in Behaviour Informatics and the multimodal study of behaviour. The study took place on Ward 35, Neonatal Intensive Care Unit, at the Royal Victoria Infirmary, Newcastle. Every year, about 60,000 preterm infants are born in the UK. These babies undergo many painful procedures as part of routine clinical care. Current pain assessment methods are not always accurate and can lead to over- or under-treatment of pain, which can harm their development.
The study aimed to develop a dataset for use in machine learning to monitor infants' comfort levels continuously. This system uses video recordings and nurses' observations to detect signs of pain more accurately and quickly. The participants in this study were infants born before 36 weeks of gestation. Only medically stable infants were included, and participation required signed parental consent.
The infants did not receive any additional treatments beyond their routine clinical care. They were recorded using a camera and microphone to capture their behaviours and responses. Bedside nurses assessed each infant using the neonatal pain, agitation, and sedation scale (N-PASS), which measures indicators like crying, behaviour state, facial expression, extremity tone, and vital signs.
Thirty-nine infants, born between 22 and 32 weeks of gestation, were recruited, resulting in 296 total recordings. The analysis is currently underway as part of a PhD project. Deep learning models are being trained to analyse each aspect of the N-PASS, including crying, behaviour state, facial expression, extremity tone, and vital signs.
The findings will be published in relevant journals and included in the PhD thesis. The trial and planned analysis was presented at the International Symposium on Paediatric Pain in Halifax, Canada, and at the Northern Neonatal Network Annual Research Conference.REC name
Yorkshire & The Humber - Leeds East Research Ethics Committee
REC reference
22/YH/0161
Date of REC Opinion
10 Aug 2022
REC opinion
Favourable Opinion