Pilot study of the eye colour & serum bilirubin level in newborns (v1)
Research type
Research Study
Full title
Pilot study of the relationship between the eye colour and serum bilirubin level and the development of a smartphone app for neonatal jaundice screening (Student Study)
IRAS ID
130932
Contact name
Judith Meek
Contact email
Sponsor organisation
University College London
Research summary
Neonatal jaundice is a common condition among newborn infants, caused by an increased bilirubin level in the blood and tissues. Bilirubin is yellow in colour and jaundiced infants can therefore appear to have yellow colouration in their skin and sclera. Most jaundice is benign but an unusually high level of bilirubin (hyperbilirubinemia) can affect the newborn’s health, and in severe cases, may even cause brain damage or death. A timely identification of hyperbilirubinemia is therefore crucial. Most babies develop jaundice over the first week of their lives. However, many hospitals have the policy to send home the mother and baby, who have no other complications, soon after birth. It is therefore left to the community midwife to identify potential hyperbilirubinemia during the home visits by visually inspecting the infant’s skin colour. This requires a high level of experience and the lighting can also change the human perception of colour, affecting the midwife’s judgement. For dark skinned babies clinical assessment can be especially unreliable. The project’s aim is to develop a smartphone App to quantify the serum bilirubin level in the newborn so that midwifes and parents can have a simple way to identify hyperbilirubinemia in the community. The App will exploit the smartphone’s built-in camera to capture a photo of the infant’s eye which is less susceptible to the influence of other pigments such as melanin and haemoglobin than skin. With an automatic algorithm, the sclera will be identified and its base colours extracted to form a new colorimetric measure as an estimate of the serum bilirubin level. Up to 200 newborn term infants of different ages will be recruited in the neonatal unit and outpatient department of UCL Hospitals. With these data, a new colorimetric measure to predict serum bilirubin level will be developed and used in the App.
Lay summary of study results
Study 1: UCLH Nikon Study The first clinical dataset used in this investigation was collected at University College London Hospital (UCLH). A digital camera was used to capture images of the sclera and forehead of
133 newborns. The aim was to investigate the potential of digital imaging for estimating the severity of jaundice, and to compare the sclera and skin as sites of measurement. Ethical Approval was obtained from the London - City Road and Hampstead NHS Research Ethics Committee. No adverse events were reported.
Subjects that required a blood test were recruited from the outpatient clinic. Many of the subjects were referred by a visiting midwife during a routine check-up, while others were attending appointments to monitor ongoing physiological jaundice. There were no exclusion criteria based of postnatal or gestational age. Babies who had received phototherapy within the preceding 24 hours were ineligible. The images were captured in the Advanced Neonatal Nurse Practitioner (ANNP) clinic of UCLH, and the blood test followed within twenty minutes to limit the possibility of a change in bilirubin level in the intervening period. Total serum bilirubin (TSB) was determined from a blood sample obtained via heel prick. The sample was spun in a centrifuge before insertion into a point-of-care bilirubinometer.
Images were captured using a Nikon D3200 24.2-megapixel DSLR camera with a 60mm macro lens. ISO was fixed at 1600, while exposure time and aperture control were set automatically by the camera. Images were saved in Nikon’s raw file format, NEF (Nikon Electronic Format). The sclera was brought into focus manually. The lighting in the ANNP clinic was not controlled beyond ensuring that the fluorescent ceiling lighting was switched on.
Analysing this dataset, Leung et al. (2015) showed the sclera-region RGB values gave a stronger correlation with TSB than did the forehead-region RGB values. This study used a quadratic polynomial model. Based on the same dataset, Leung et al. (2019) later showed that a simple linear model based on chromaticity could achieve similar correlation with TSB.
Study 2: UCLH Smartphone Study
A clinical study was carried out in the UCLH Neonatal Care Unit and Postnatal Ward between January 2017 and February 2018. No adverse events were reported.
The aim was to investigate the accuracy of image-based jaundice estimation via the sclera using a smartphone. This study used the front-facing camera of an LG Nexus 5X smartphone, with illumination provided by the screen of the smartphone itself.
The pictures were taken at whatever time was most convenient during the appointment so as not to disrupt the assessment or treatment of the baby. Often, the data was collected before the baby underwent a blood test, while he or she was lying supine on the assessment table.
Babies were found to be more compliant before the blood test, as the heel prick to draw blood is a painful procedure. In other cases, the images were captured with the baby in the arms of a parent.
Around the time of feeding babies open their eyes, making data collection more straightforward. The pictures were always taken within 20 minutes of the blood draw so the TSB reading would be accurate at the time of image capture. Image capture involved holding the phone approximately 10-20cm from the face. Multiple image pairs were captured for each subject to ensure at least one pair was usable. The user would aim to ensure that the image was not blurred and that the sclera was visible. However, due to the clinical setting this was not always possible.
The TSB measurement began with a blood draw using a heel lance. A few drops were collected using a capillary tube, which was promptly sealed with resin. A centrifuge spun the tube to separate the serum from the hematocrit, and the serum was analysed using a bilirubinometer to give a value for the total serum bilirubin (TSB). This method allowed a result at the point of care. Sometimes this method was not available, and a sample was instead sent for laboratory testing, which is a longer process. Only measurements taken using the centrifuge and bilirubinometer method were used for this study.
A total of 51 babies were imaged using the application (16 female, 35 male, median postnatal age 11 days). All but four of the 51 image sets were captured in the same room (the Advanced Neonatal Nurse Practitioner (ANNP) Clinic of the UCLH Neonatal Ward). For eight of the subjects, not enough sclera was visible to identify a usable region of pixels in the sclera. For another two subjects, results from the point-of-care bilirubinometer were not available. 41 subjects remained for analysis.
Based on this dataset, a journal article is being prepared investigating ambient-subtracted sclera chromaticity for smartphone-based jaundice screening.
References
Leung, T.S., Kapur, K., Guilliam, A., Okell, J., Lim, B., MacDonald, L.W., and Meek, J.:
‘Screening neonatal jaundice based on the sclera color of the eye using digital photography’, Biomedical Optics Express, 2015, 6, (11), pp. 4529-4538 Leung, T.S., Outlaw, F., MacDonald, L.W., and Meek, J.: ‘Jaundice Eye Color Index (JECI):
quantifying the yellowness of the sclera in jaundiced neonates with digital photography’, Biomedical Optics Express, 2019, 10, (3), pp. 1250-1256REC name
London - Bromley Research Ethics Committee
REC reference
13/LO/1853
Date of REC Opinion
28 Jan 2014
REC opinion
Further Information Favourable Opinion