Imaging software instantly diagnoses, treats jaundice
Engineers from the University of South Australia (UniSA) and Middle Technical University have designed imaging software that can diagnose jaundice in the blink of an eye, automatically turn on an LED light to counteract it and send the diagnosis via SMS to the carer. Their breakthrough has been published in the journal Designs.
Jaundice is a common condition in newborns, especially premature babies, where there is an overload of an orange-yellow pigment called bilirubin in the bloodstream. It normally resolves quickly when the baby’s liver is mature enough to remove it from the body, but in severe cases can be treated by phototherapy, whereby fluorescent blue light is used to break down the bilirubin in the baby’s skin.
UniSA remote sensing engineer Professor Javaan Chahl says jaundice is particularly prevalent in developing countries, where there often isn’t the equipment or trained medical staff to effectively treat it; as a result, an estimated 75,000 children are currently living with brain dysfunction worldwide due to complications from jaundice. The newly developed imaging software is designed to counteract this.
“Using image processing techniques extracted from data captured by the camera, we can cheaply and accurately screen newborns for jaundice in a non-invasive way, before taking a blood test,” Prof Chahl said.
“When the bilirubin levels reach a certain threshold, a microcontroller triggers blue LED phototherapy and sends details to a mobile phone.
“This can be done in one second, literally, which can make all the difference in severe cases, where brain damage and hearing loss can result if treatment is not administered quickly.”
Researchers tested the system in an intensive care unit in Mosul, Iraq, on 20 newborns diagnosed with jaundice. A second dataset captured 16 images of newborns — five healthy and the remainder jaundiced. The system was also successfully tested on four other manikins with white and brown skin colours, with and without jaundice pigmentation.
“Previous research using sensors to find a non-invasive way to detect jaundice has fallen short,” said Prof Chahl. “Methods trialled have been unreliable, costly, inefficient and in some cases caused infections and allergies where sensors needed skin contact.
“Our system overcomes these obstacles by immediately detecting jaundice based on a novel digital representation of colour which allows high diagnostic accuracy at a relatively low cost. It could be widely used in hospitals worldwide and medical centres where laboratory facilities and trained medical staff are not available.”
AI systems can diagnose prostate cancer biopsies with the same level of accuracy as specialist...
The system allows automatic analysis of biomedical videos captured by microscopy in order to...
The technique uses an AI process known as deep convolutional neural network, an algorithm...