Dr. Dan Ponticiello, 43, and Dr. Gabriel Gomez, 40, are admitting a patient with coronavirus infection (COVID-19) in the COVID-19 ICU at Providence Mission Hospital in Mission Viejo, California, Jan. 8 , 2021.
Lucy Nicholson | Reuters
Facebook’s artificial intelligence researchers say they have developed software that can predict the likelihood of a Covid patient declining or needing fixed oxygen. on X-rays of the chest.
Facebook, which worked with academics at NYU Langone Health’s predictive analysis unit and radiation research department, says the software could help doctors send home at-risk patients premature, and also helping hospitals plan for oxygen demand.
The 10 researchers involved in the study – five from Facebook AI Research and five from NYU School of Medicine – said they have developed a total of three machine learning “models”, all of which are slightly different.
One attempts to predict patient degeneration 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 extra oxygen (if any) a patient may need.
“Our model using continuous breast X-rays can predict up to four days (96 hours) in advance if a patient may require more intensive care solutions, typically performing better not predicted by human experts, “the authors said in a blog post published Friday.
William Moore, professor of radiation at NYU Langone Health, said in a statement: “We have been able to show that, by using this AI algorithm, serial breast radiographs can predict the need for increased care. the patient with Covid-19. “
He said: “As Covid-19 remains a key public health issue, the ability to predict patient need for increased care – for example, admission to an ICU – will be critical for hospitals.”
To learn how to predict, two data sets of non-Covid patient breast X-rays were fed into the AI system and a data set of 26,838 breast X-rays from 4,914 Covid patients.
The researchers said they used an AI technique called “contrast momentum” to train a neural network to extract information from chest X-ray images. A neural network is a computer system that is clearly stimulated by the human brain that can recognize patterns and identify relationships between large amounts of data.
The research was published by Facebook this week but experts have already questioned how effective the AI software can be in action.
“From a machine learning perspective, one had to examine how well this translates into new, unseen data from different hospitals and patient numbers,” said Ben Glocker, who studies machine learning. for images at Imperial College London, via email. “From my skim reading, it seems like all the data (training and testing) comes from the same hospital.”
Researchers on Facebook and NYU said: “These models are not results, but research solutions, aimed at helping hospitals in the coming days and months with facility design. hospitals have their own unique data, which often lacks the computing power necessary to train deep learning models from the outset. “
“We are open to our pre-designed models (and publish our results) so that hospitals with very little computing resources can customize the models using their own data,” they said.