The University of Illinois Chicago (UIC) is developing an artificial intelligence system that will aim to protect US Navy divers from harmful waterborne pathogens.
Researchers have been awarded a two year grant of $725,000 from the US Office of Naval Research for the project, which could help to improve understanding of the health risks that exist for sailors when conducting underwater duties, such as fleet maintenance and repairs or research and rescue missions.
Current water testing technologies typically rely on lab-based analysis of samples and scientists knowing which microbes to screen, but the exact condition of coastal water at specific times is hard to predict due to dynamic weather, currents, water temperatures and pollution.
“By the time a water sample arrives at a lab and is tested, the conditions may have changed,” said co-principal investigator Dr Samuel Dorevitch, associate professor of environmental and occupational health sciences at UIC’s School of Public Health.
“If navy divers had real-time information, they could select the best protective equipment, dive duration and take other measures to prevent the various health issues, like heat stress or gastrointestinal, skin and respiratory infections that may result from microbes in water.”
Co-principal investigator Isabel Cruz, professor of computer science at the College of Engineering, added that AI can offer a way to ‘synthesise a vast amount of information quickly for a specific calculation’.
Researchers hope to develop a system that could be used in any location by divers to analyse water conditions through a combination of user-provided and web-based information and human data, such as divers’ age, health and the diving team’s size.
“This project is both exciting and challenging because of its multidimensionality,” said Cruz. “We hope to pull information from many sources that offer different types of data, and we will have to integrate data that are quite complex, heterogeneous, and often without metadata.”
The AI and machine learning methods will be built in stages, Cruz said, adding that the system could prove ‘remarkable’ if it can be taught to reliably and accurately prioritise all data for risk prediction.