GE Healthcare is adopting an Artificial Intelligence (AI) system that detects COVID-19 symptoms and assists front line workers in the placement of ventilator tubes. DeepCOVID-XR is designed with algorithms that can read x-rays and quickly screen patients in the hospital. The product is included in GE Healthcare's Critical Care Suite and has the potential to screen patients for signs of health risks unrelated to COVID-19 as well.
The product is attached to an x-ray scanner and can spot signs of COVID-19 ten times faster than radiologists and is up to 6% more accurate. DeepCOVID-XR was tested on over 17,000 x-ray images and collected patterns of symptoms including hazy and inflamed lungs and fluid filled air sacs. Much of the collected data reflected symptoms of other illnesses such as pneumonia and heart failure, however, with the screening, doctors are informed of other possible treatments.
During testing, DeepCOVID-XR produced results faster than a group of trained radiologists. DeepCOVID-XR processed 300 x-rays in 18 minutes while radiologists took up to three hours. The product's accuracy rate was 82% while the radiologists had 76% to 81% accuracy.
DeepCOVID-XR not only detects symptoms but can help workers determine the correct placement for ventilation tubes. An estimate of one in four patients with ventilation tubes outside of the hospital have a misplaced endotracheal tube and up to 45% of individuals in intensive care received intubation connected to a ventilator. Misplacement can lead to lethal consequences such as a collapsed or overinflated lung. DeepCOVID-XR locates the tube inserted in the chest automatically during the x-ray and offers measurements of its placement within the windpipe. Through the scan, it also provides notification of possible dangers to workers and addresses the need for isolation.
The COVID-19 pandemic has left frontline workers overwhelmed and more prone to error with the increased number of patients. DeepCOVID-XR aims to offer artificial intelligence to assist workers with fast and accurate data for treating patients. The researchers from Northwestern University have published the algorithm enabling others to train the system with new data for beneficial upgrades in hopes of releasing the service to clinics.
