Facebook claims its AI can predict four days in advance if a coronavirus patient’s condition will worsen – just by looking at a single chest X-ray
- Facebook and NYU have built AI to predict developments among COVID patients
- IT was trained with thousands of chest X-rays with and without COVID
- The software is based on a series of three machine learning models
- One predicts deterioration in one scan and the other uses a series of scans
- The third model can predict whether supplemental oxygen will be required
Facebook claims to have designed software that can predict whether a coronavirus patient’s health will deteriorate or require oxygen by simply scanning their chest X-rays.
In collaboration with New York University (NYU), the social media company says the system can calculate such developments in four days.
Together, they have built three machine learning models to help doctors better prepare as cases around the world continue to increase.
One model is designed to predict deterioration using a single chest X-ray, another does the same thing but through a series of X-rays, and the third uses an X-ray to determine if and how much supplemental oxygen a patient needs.
Scroll down for video
Facebook and NYU built three machine learning models to help doctors better prepare as cases continue to increase around the world. One model is designed to predict deterioration using a single chest X-ray, another does the same but through a series of X-rays
The coronavirus has held the world tight for more than a year, starting in Wuhan, China.
Worldwide, there have been more than 93 million cases and about two million people have died from the virus.
However, the US has suffered with more than 23 million cases and more than 389,000 deaths.
As cases continue to increase, scientists and technology companies are working on innovative ways to cope with symptoms before they take a deadly turn.
The third uses an X-ray to determine if and how much supplemental oxygen a patient needs. The AI can make predictions based on the X-rays four days in advance
Facebook and NYU each have five researchers who worked on the coronavirus prediction software.
‘Our model using sequential chest X-rays can predict up to four days (96 hours) in advance whether a patient may need more intensive care solutions, better overall than human experts’ predictions,’ the authors said in a Friday. published blog post.
The team first ‘pre-trained’ the model with a public database of chest X-rays on MIMIC-CXR-JPG and CheXpert.
The system was then trained on NYU coronavirus X-rays, which contained a total of 26,383 images of 4,914 patients, allowing the models to label a patient’s deterioration by 24, 48, 72, or 96 hours of the scan in matter.
Based on reader surveys we conducted with NYU Langone radiologists, our models using X-ray sequences outperformed human experts in predicting ICU needs and mortality predictions, and longer-term overall predictions of adverse events (up to 96 hours). ), ‘the team shares in the blog.
“Being able to predict whether a patient needs oxygen would also be a first, and could help hospitals decide how to allocate resources in the weeks and months ahead.”
Facebook claims to have designed software that can predict whether a coronavirus patient’s health will deteriorate or require oxygen just by scanning the chest X-rays
William Moore, MD, professor of radiology at NYU Langone Health, shares in a statement, “ We have been able to demonstrate that using this AI algorithm, serial chest radiographs can predict the need for escalation of care in patients with COVID. -19. ‘
“Since COVID-19 is still a major public health problem, it will be essential for hospitals to be able to predict whether a patient needs more care, for example in an ICU.”