Fb creates AI to foretell the chance of Covid signs getting worse
Dr. Dan Ponticiello (43) and Dr. Gabriel Gomez, 40, intubate a patient with coronavirus disease (COVID-19) in the COVID-19 intensive care unit at the Providence Mission Hospital in Mission Viejo, California on January 8, 2021.
Lucy Nicholson | Reuters
Artificial intelligence researchers at Facebook claim they have developed software that can use the x-rays of their chest to predict the likelihood that a Covid patient will deteriorate or need oxygen.
Facebook, which worked on research with scientists from the Predictive Analytics Unit and the Radiology Department at NYU Langone Health, says the software could help doctors avoid sending high-risk patients home early and help hospitals meet oxygen needs to plan.
The 10 researchers involved in the study – five from Facebook AI Research and five from the NYU School of Medicine – said they had developed a total of three “models” of machine learning, all of which are slightly different.
One tries to predict the patient’s deterioration from a single chest x-ray, another does the same thing with a series of x-rays, and a third uses a single x-ray to predict how much additional oxygen (if any) a patient may need.
“Our sequential chest x-ray model can predict up to four days (96 hours) in advance whether a patient may need more intensive care solutions that generally exceed predictions by human experts,” the authors said in a blog post published Friday.
William Moore, professor of radiology at NYU Langone Health, said in a statement, “We were able to show that this AI algorithm can use serial chest x-rays to predict the need for escalation of care in patients with Covid-19.”
He added, “As Covid-19 continues to be a major public health issue, the ability to predict a patient’s need for increased care – such as ICU admission – will be vital for hospitals.”
To learn how to make predictions, the AI system was fed two sets of non-Covid x-rays of the patient and a data set of 26,838 chest x-rays from 4,914 Covid patients.
The researchers said they used an AI technique called “pulse contrast” to train a neural network to extract information from X-rays of the chest. A neural network is a computer system vaguely inspired by the human brain that can recognize patterns and identify relationships between huge amounts of data.
The study was published by Facebook this week, but experts have already questioned how effective the AI software can be in practice.
“From a machine learning perspective, we need to investigate how well it can produce new, invisible data from different hospitals and patient groups,” said Ben Glocker, who studies machine learning for imaging at Imperial College London, via e- Mail. “From my skimming reading, it can be seen that all of the data (training and testing) came from the same hospital.”
The researchers from Facebook and NYU said, “These models are not products, but research solutions that should help hospitals resource planning in the coming days and months. While hospitals have their own data sets, they often do not have the computing power needed to To train deep learning models from scratch. “
“We are making our pre-trained models available (and publishing our results) as open-sourced products so that hospitals with limited computing resources can optimize the models based on their own data,” they added.