Dr. Dan Ponticiello, 43, and Dr. Gabriel Gomez, 40, intubate a patient with coronavirus disease (COVID-19) at ICU COVID-19 at Providence Mission Hospital in Mission Viejo, California, January 8, 2021.
Lucy Nicholson | Reuters
Facebook’s artificial intelligence researchers say they have developed software that can predict the likelihood that a Covid patient will deteriorate or need oxygen based on his chest X-rays.
Facebook, which worked with academics from NYU Langone Health’s predictive analytics unit and radiology department on the research, says the software can help doctors avoid sending patients at risk home too early, while helping hospitals to 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 developed three machine learning “models” in total, which are slightly different.
One tries to predict the patient’s deterioration based on a single chest X-ray, another does the same with a sequence of X-rays, and a third uses a single X-ray to predict how much supplemental oxygen (if any) a patient can need .
“Our model using sequential chest radiographs can predict up to four days (96 hours) in advance if a patient may need more intensive care solutions, often exceeding the predictions of human experts,” said the authors in a blog published on Friday -market.
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 chest radiographs can predict the need for intensified care in patients with Covid-19.”
He added: “As Covid-19 remains a major public health problem, the ability to anticipate the need to increase patient care – for example, admission to the ICU – will be essential for hospitals.”
To learn how to make predictions, the AI system was fed two sets of chest radiographs data from non-Covid patients and one set of 26,838 chest radiographs from 4,914 Covid patients.
The researchers said they used an AI technique called “moment contrast” to train a neural network to extract information from chest X-ray images. A neural network is a computer system loosely 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 questioned how effective AI software can be in practice.
“From a machine learning perspective, it would be necessary to study how this translates into new, invisible data from different hospitals and patient populations,” said Ben Glocker, who researches machine learning for images at Imperial College London, by email . “From my cursory reading, it looks like all the data (training and testing) comes from the same hospital.”
The Facebook and NYU researchers said, “These models are not products, but research solutions, designed to help hospitals in the coming days and months with resource planning. Although hospitals have their own unique data sets, many sometimes they don’t have the computing power needed to train deep learning models from scratch. “
“We are making our pre-trained models available (and publishing our results) so that hospitals with limited computing resources can adjust the models using their own data,” they added.