Computer Vision Based Jaundice Monitoring for Smart Low-Cost Phototherapy Light System

Document Type

Conference Proceeding

Publication Date



This paper describes the development and enhancements of the Smart Low-cost Phototherapy Light System (Smart LPLS) at Ateneo Innovation Center (AIC) and partners. The developed Smart LPLS has custom-built LED panels that can output an irradiance of up to 83.89 μW/cm2/n; 2.7 times stronger than the minimum required of 30 μW/cm2/nm for Neonatal Jaundice treatment. The integrated monitoring system uses an ESP32 camera module for capturing and sending images and a Raspberry Pi that enables Near Cloud storage, processing, and rendering capabilities. Using a modified version of Kovac's rules for skin pixel identification and K-means clustering for color quantization proved to be effective in extracting the dominant colors present within a skin region. Extracted colors are checked for yellow pigmentation based on their b∗ axis value on the CIE L*a*b∗ color space, which indicates yellowness (-) or blueness (+). This enables the tracking of increase and decrease in yellow pigmentation over time, with a configurable threshold for sending alerts through email at perceived dangerous levels. The Smart LPLS offers a cost-effective and efficient alternative to commercial phototherapy units, with the additional functionality of an integrated smart monitoring system.