Cambridge’s Addenbrooke’s Hospital in collaboration with NVIDIA and 20 other hospitals around the world has used an Artificial Intelligence algorithm and Federated Learning to predict the oxygen needs of Covid-19 patients in the first days of therapy.
FEDERATED LEARNING: FEDERATED LEARNING FOR ALGORITHMIC TRAINING
The AI algorithm employed by Addenbrooke’s Hospital in Cambridge was based on a machine learning technique called Federated Learning.
With Federated Learning we mean a machine learning technique in algorithmic training in which, unlike traditional learning systems that require the sharing and maintenance of data on a single database or server, the data is not shared but anonymized and encrypted for train the algorithm (1).
See also AI Algorithm Can Diagnose Covid-19 And Other Diseases While Keeping Patient Privacy Intact.
In this way, the data can also be hosted on different databases and scattered around the world: in this way, privacy is guaranteed because it is not necessary to transfer data from the device in which it was initially collected.
THE CASE OF THE ADDENBOOK’S HOSPITAL IN CAMBRIDGE
Using this technique, the Addenbrooke’s Hospital algorithm analyzed x-rays and other (anonymous) data of patients presenting with Covid-19 symptoms.
Once the algorithm was trained, the analyzes carried out were put together to create an AI tool capable of predicting the amount of oxygen required for hospitalized patients anywhere in the world.
EXAM: THE STUDY BASED ON FEDERATED LEARNING
The study called EXAM (EMR CXR AI Model) and published in Nature Medicine, was tested in several hospitals on 5 continents: within 24 hours of the patient’s admission to the hospital, the algorithm predicted, with a sensitivity of 95% and an accuracy of 85%, the oxygen requirement of each patient.
Professor Fiona Gilbert, honorary radiologist consultant at Addenbrooke’s Hospital, stressed that Federated Learning is a great transformative element in AI innovation, especially in the clinical setting.
“Federated learning has enabled researchers to collaborate and set a new standard for what we can do globally, using the power of AI,” said Dr Mona G Flores, NVIDIA’s Global Head for Medical AI. . “This will advance AI not only in healthcare, but in all sectors that seek to build robust models without sacrificing privacy.”
Professor Gilbert added: “Creating software that matches the diagnoses of our best radiologists is complex. The more we are able to securely integrate data from multiple sources using federated learning, and the more space we need to innovate, the faster academics will be able to turn these goals into reality. “
Resources:
https://www.nature.com/articles/s41591-021-01506-3
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8609899/
(1) https://ai.googleblog.com/2017/04/federated-learning-collaborative.html