Dr. Dan Ponticiello (43) and dr. Gabriel Gomez (40) intubates a coronavirus (COVID-19) disease in the COVID-19 ICU at Providence Mission Hospital in Mission Viejo, California, on January 8, 2021.
Lucy Nicholson | Reuters
Artificial intelligence researchers at Facebook claim to have developed software that could predict the likelihood of a Covid patient deteriorating or needing oxygen based on their X-rays on the chest.
Facebook, which worked with academics from NYU Langone Health’s unit for predictive analysis and the radiology division on the research, says the software can help doctors prevent patients at risk from being sent home too early while also hospitals can help plan oxygen demand.
The ten researchers involved in the study – five from Facebook AI Research and five from the NYU School of Medicine – said they developed a total of three models for machine learning, all of which differed slightly.
One tries to predict patient deterioration based on a single chest X-ray, another does the same with a series of x-rays, and a third uses a single X-ray to predict how much supplemental oxygen (if any) a patient needs. .
“Our model that uses consecutive chest X-rays can predict up to four days (96 hours) in advance if a patient needs more intensive care solutions, which are usually better than the predictions of human experts,” the authors said in a blog post published Friday.
William Moore, a professor of radiology at NYU Langone Health, said in a statement: “We were able to show that using this AI algorithm, serial radiographs can predict the need for increased care in patients with Covid-19. . “
He added: “As Covid-19 continues to be a major public health problem, the ability to predict a patient’s need for increased care – for example, admission to ICU – is essential for hospitals.”
To learn how to make predictions, the AI system fed two sets of non-Covid patient chest X-rays and a data set of 26,838 chest X-rays of 4914 Covid patients.
According to the researchers, they used an AI technique called ‘momentum contrast’ to train a neural network to take information out of the chest. A neural network is a computer system vaguely inspired by the human brain that can detect patterns and recognize relationships between large amounts of data.
The research was published by Facebook this week, but experts have already asked how effective the AI software can be in practice.
“From a machine learning perspective, one has to study how well it translates into new, unseen data from different hospitals and patient populations,” said Ben Glocker, who conducts email learning on imaging machine learning at Imperial College London. . “From my reading, it appears that all data (training and testing) comes from the same hospital.”
The researchers from Facebook and NYU said: ‘These models are not products, but rather research solutions, which are meant to help hospitals in the days and months with resource planning. Although hospitals have their own unique datasets, they often do not have the computing power needed to train deep learning models from scratch. “
“We are open-sourcing our pre-trained models (and publishing our results) so that hospitals with limited computer resources can refine the models using their own data,” they added.