Bilirubin estimates from smartphone images of newborn infants’ skin correlated highly to serum bilirubin levels

Abstract

AIM: Neonatal jaundice is an important cause of morbidity and mortality, and identifying the condition remains a challenge. This study evaluated a novel method of estimating bilirubin levels from colour-calibrated smartphone images. METHODS: A cross-sectional prospective study was undertaken at two hospitals in Norway from February 2017 to March 2019, with standardised illumination at one hospital and non-standardised illumination at the other hospital. Healthy term-born infants with a normal birthweight were recruited up to 15 days of age. The main outcome measures were bilirubin estimates from digital images, plus total bilirubin in serum (TSB) and transcutaneous bilirubin (TcB). RESULTS: Bilirubin estimates were performed for 302 newborn infants, and 76 had severe jaundice. The correlation between the smartphone estimates and TSB was measured by Pearson's r and was .84 for the whole sample. The correlation between the image estimates and TcB was 0.81. There were no significant differences between the hospitals. Sensitivity was 100%, and specificity was 69% for identifying severe jaundice of more than 250 µmol/L. CONCLUSION: A smartphone-based tool that estimated bilirubin levels from digital images identified severe jaundice with high sensitivity and could provide a screening tool for neonatal jaundice.