Ai-based test to help newborns from shock
NEW DELHI: A team of Indian researchers has developed a thermal imaging and artificial intelligence-based test that predicts setting in of haemodynamic shock -- insufficient oxygen supply to organs leading to multi-organ failure -- in children even 12 hours before doctors can clinically diagnose it. The detection of shock in time can prevent organ failure and save lives.
The shock in critically ill patients can set in due to various conditions, including heavy internal bleeding and severe blood infection called sepsis that leads to restricted blood flow and a resulting lack of oxygen supply to organs and tissues, causing death. Sepsis is one of the top three causes of neonatal (till end of first month) deaths in India, with the neonatal mortality rate (NMR) being 28 per 1000 live births. The new research, published in the journal Nature’s Scientific Report on January 14 this year, has been tested at the paediatric intensive care unit (PICU) of Delhi’s All India Institute of Medical Sciences (AIIMS) for about a year and a half with a success rate of of 69% in detecting shock 12 hours in advance.
Children in the age group of 10 days to 15 years in the AIIMS PICU were tested for the study titled “Predicting Haemodynamic Shock from Thermal Images using Machine Learning”. The Indraprastha Institute of Information Technology, Delhi, (IIIT-D) and AIIMS collaborated on the five-year project that has been funded by Welcome Trust and Department of Biotechnology (DBT) and will go on till 2020.
In about 60%-90% of ICU cases, sepsis may develop, and about 30% tend to get into shock, of which 30% don’t make it.
“Since shock is a leading killer of children in ICUS, the theme of my project was how to detect it early so that treatment can be started soon enough. We developed a tool that combines reading of temperature patterns over the body using artificial intelligence,” said Dr Tavpritesh Sethi, a faculty at IIIT-D and a Welcome Trust fellow at AIIMS.
The test is non-invasive and does not require having contact with a patient, avoiding the risk of further infection.
“Patients in the ICU should come in contact with outside articles as little as possible to cut further infection risk. This test would require clicking of images with a special camera attached to a smartphone. The images stored in the cell phone will be transferred to a computer or uploaded on a server for analyses,” said Sethi.
On average, the researchers clicked five images of the feet and abdomen of a patient but that was for research. One image is generally enough to get an indication of shock setting in.
As part of the project, the researchers will be expanding the test for continuous monitoring at AIIMS PICU.
“We will also be doing pilot trials from this August until July 2020 to get an idea of the number of lives saved through this. Right now it is for children, but it can be updated for adults,” said Sethi.