Zhang, Yifan; (2024) Application of a Tissue Colour Model in Smartphone Imaging of Neonatal Jaundice. Doctoral thesis (Ph.D), UCL (University College London).

Text Zhang_10201061_thesis.pdf Access restricted to UCL open access staff until 1 January 2026. Download (6MB)

Abstract

Over 60% of newborns develop jaundice in their first week of life. Jaundice, which introduces yellow discolouration in the newborns’ skin and sclera, must be carefully monitored to avoid possible neurological damage. It is responsible for over 100,000 deaths every year. Over 70% of the cases are attributed to low and middle income countries (LMICs) in Africa and South Asia, where healthcare resources are in shortage. There is a need for an affordable jaundice screening technique. This thesis investigates a smartphone camera-based technique for jaundice screening based on the colour of newborns’ sclera. As the core part of the screening technique, a tissue colour model was developed by incorporating both the camera imaging theory and the tissue diffuse reflectance spectroscopy theory. Using forward modelling, we can predict the tissue colour in the visible spectrum from the bilirubin level, while the bilirubin level could be estimated from the measured tissue colour through inverse modelling. We have introduced two different methods for estimating bilirubin level from tissue colour. The first method, based on a regression model between the total serum bilirubin (TSB) measured by blood tests and sclera chromaticity measured by a digital camera, provides a predicted bilirubin which we term as sclera-conjunctival bilirubin (SCB_{reg}). The second method, based on the newly developed tissue colour model, estimates the SCB_{tis} using sclera chromaticity. The tissue colour model was validated on clinical data collected from Ghana, as part of the Detecting Jaundice in African Newborn (DJAN) study. In a high-quality image database (n = 438), the performance of the two models was similar. For detecting significant jaundice (TSB> 250\mu mol/L), while the sensitivity and specificity of SCB_{reg} were 1 and 0.93, those of SCB_{tis} were 1 and 0.95, respectively. Although the smartphone camera-based technique performs well as a screening technique providing a binary decision (below or above a threshold), it is not accurate to estimate TSB directly. This thesis also investigated the main factors that limit the accuracy of TSB estimation, including the minimum detectable bilirubin concentration change for a given colour depth of a camera, optimizing the quantification of yellowness with coloured illumination, estimating bilirubin in the presence of blood, and optimal calibration of a camera.

Downloads since deposit

Loading...

4Downloads

Download activity - last month

Export as XMLJSONCSV

Loading...

Download activity - last 12 months

Export as CSVJSONXML

Loading...

Downloads by country - last 12 months

Export as XMLJSONCSV

Loading...

Archive Staff Only

View Item