Non-invasive jaundice detection system using colour card technique

Jaundice is described as yellow discoloration of the skin and other tissues of a newborn infant and it is conjunctiva due to the increase of bilirubin up to 5 mg/dl or 85 𝛍mol /l. In Malaysia, a survey made by Ministry of Health Malaysia proves that about 75% ofnewborns had jaundiced in the first week of life. Current jaundice assessment method have their own drawbacks. The traditional method (Kramer's Rule) is not convenientsince there are limits to the accuracy of visual estimation and may lead to human error. Thus, a blood test taken from baby's heel remains to be necessary but it leads to traumatization for the baby as well as their parents. Moreover, the available jaundice meter in the market is expensive. This jaundice detection using color card demonstrated in this research shows a high potential to give positive impact on the medical and healthcare fields. This research work proposed an automated jaundice (hyperbilirubinemia) detection system and the suggested treatment based on the color card technique. This research does not involve any subject or infant. It acquires data by capturing different type of color shades using spectral reflectance method. It represent random bilirubin color in patient's body, the human skin color represented by Fitzpatrick shades, and bilirubin concentration color card using silicon photodiode sensor. The data gained are transferred wirelessly to a gateway and subsequently to a computer to perform computation and analysis. The input data were validated using Anderson-Darling normality and compared with the reference card using linear regression and Pearson's correlation coefficient analysis to predict the particular bilirubin value and treatment needed for each input sample tested. Percentage error analysis is used to verify and validate the experimental device. The final outputs are shown in a graphical user interface on medical staffs PC and can be viewed only by the authorized personnel. Based on the results of linear relationship and correlation between input sample and reference data, it shows high percentage result (R²) between 86.4% to 99.4%, and correlation value (r) 0.918 to 0.997. Result of highest percentage error (% error) shows value of 9.90% and verify that this device is reliable. Thus, it can be concluded that the color card technique is reliable and can be used to detect and predict the jaundice level for any type of skin color.