Researchers from across the world have developed a tool that uses artificial intelligence (AI) to predict how much oxygen a Covid-19 patient will need during hospital admissions.
Several hospitals across five continents were used to test the AI tool’s accuracy.
It was found to be 95 percent sensitive and 88 percent specific in predicting whether oxygen would be needed within 24 hours of patient arrival in an emergency department.
In a study published Thursday in the journal Nature Medicine, research outcomes from around 10,000 patients with Covid-19 were analyzed.
Using federated learning, experts analysed chest radiographs and electronic health data from patients with symptoms of Covid-19.
As a means of maintaining strict patient confidentiality, all patient data has been anonymized and given a unique algorithm within each hospital. No data has been shared or left the hospital.
A machine learning algorithm was developed from the data and the analysis was used to construct the AI tool.
“Federated learning has transformative power to bring AI innovation to the clinical workflow,” said Professor Fiona Gilbert, from the University of Cambridge in the UK, who led the study.
“Usually in AI development, when you create an algorithm on one hospital’s data, it doesn”t work well at any other hospital,” said study first author Ittai Dayan, from Mass General Bingham in the US.
Researchers constructed a generalizable model using objective, multimodal data from different continents in order to help frontline clinicians around the world.
In just two weeks of AI ‘learning,’ this study successfully predicted high-quality results from collaborations in North and South America, Europe, and Asia.
“Federated Learning allowed researchers to collaborate and set a new standard for what we can do globally, using the power of AI,” said Mona G Flores, Global Head for Medical AI at healthcare technology company NVIDIA.
“This will advance AI not just for healthcare but across all industries looking to build robust models without sacrificing privacy,” Flores said.