[ad_1]
Researchers at Facebook and NYU Langone Health have created AI models that analyze x-rays to predict the course of a COVID-19 patient’s condition.
The team says their system can predict whether a patient may need more intensive care resources up to four days in advance. They believe hospitals could use it to anticipate demand for resources and avoid sending at-risk patients home too soon.
Their approach differs from most previous attempts to predict the deterioration of COVID-19 by applying machine learning techniques to x-rays.
These typically use supervised training and single period images. This method has shown promise, but its potential is limited by the tedious process of manually labeling data.
[Read: How Netflix shapes mainstream culture, explained by data]
These limitations have led researchers to resort to self-supervised learning instead.
They first pre-trained their system on two public x-ray datasets, using a self-supervised learning technique called Momentum Contrast (MoCO). This allowed them to use a large amount of non-COVID x-ray data to train their neural network to extract information from the images.
Predicting the deterioration of COVID-19
They used the pre-trained model to create classifiers that predict whether a COVID-19 patient’s condition is likely to worsen. They then refined the model with an extended version of the NYU COVID-19 dataset.
This smaller data set of about 27,000 x-ray images from 5,000 patients were given labels indicating whether the patient’s condition had deteriorated within 24, 48, 72 or 96 hours of the examination.
The team built a classifier that predicts patient deterioration based on a single x-ray. Another makes his predictions using an X-ray sequence, aggregating the features of the image through a Transformer model. A third model estimates how much extra oxygen patients might need by analyzing an x-ray.
They say the use of an x-ray sequence is especially valuable because they are more accurate for long-term predictions. This approach also takes into account the evolution of infections over time.
Their study showed that the models were effective at predict intensive care needs, prediction of mortality and overall prediction of longer-term adverse events (up to 96 hours):
The performance of our multi-image model exceeded that of all single-image models. Compared to radiologists, our multi-image prediction model was comparable in its ability to predict patient deterioration and stronger in its ability to predict mortality.
The team opened up the pre-trained models so that other researchers and hospitals could refine them with their own COVID patient data – using a single GPU.
You can read the study paper on the Axiv.org preprint server.
Published January 15, 2021 – 5:00 PM UTC
[ad_2]
Source link