Using data from five continents, over 20 hospitals around the world collaborated with NVIDIAA, a pioneer in AI technology, to develop a novel AI-based approach known as federated learning. The method analyses chest X-rays and electronic health data from hospital patients with Covid symptoms using an algorithm.
The analysis was brought together to develop an AI tool — EXAM (electronic medical record (EMR) chest X-ray AI model) — that could forecast the oxygen demands of hospital Covid patients anywhere in the globe once the algorithm had "learned" from the data.
The findings, which were published in the journal Nature Medicine, revealed that it could estimate the amount of oxygen required within 24 hours of a patient's admission in the emergency room, with a sensitivity of 95% and a specificity of over 88%.
To ensure absolute patient confidentiality, all patient data were anonymized, and an algorithm was given to each institution to ensure that no data was exchanged or left its original location.
Federated Learning enables academics to collaborate and create a new global benchmark for what AI can do. This will help AI develop not just in healthcare, but in any industry that wants to construct reliable models without sacrificing privacy. The research looked at the results of about 10,000 Covid patients from all around the world.