by John R. Fischer
, Senior Reporter | December 08, 2020
Researchers at Northwestern University have developed DeepCOVID-XR, an AI platform for detecting COVID-19 from X-ray images of the lung.
The platform identified the virus about 10 times faster and 1% – 6% more accurately than a team of specialized thoracic radiologists.
"By analyzing images that were correctly and incorrectly classified by the algorithm as well as heatmaps of important features for predicting COVID-19 generated by the algorithm, we were able to determine that DeepCOVID-XR homes in on classic features of COVID-19-associated pneumonia on chest X-ray including bilateral, patchy consolidations (fluid/inflammation in the lungs), especially along the periphery and lower lobes," Dr. Ramsey Wehbe, a cardiologist and postdoctoral fellow in A.I. at the Northwestern Medicine Bluhm Cardiovascular Institute, told HCB News.
Numed, a well established company in business since 1975 provides a wide range of service options including time & material service, PM only contracts, full service contracts, labor only contracts & system relocation. Call 800 96 Numed for more info.
The system could screen patients admitted for non-COVID-19 reasons and identify asymptomatic carriers, allowing them to isolate sooner. While it does not confirm cases of COVID-19, it can flag a suspicious patient and speed up triage before the test results come back, which can take hours or days.
The researchers developed, trained and tested the system with 17,002 chest X-ray images, which they say is the largest published clinical data set of chest X-rays during COVID-19 for training an AI system. Of these images, 5,445 came from positive patients.
While each of five experienced cardiothoracic fellowship-trained radiologists spent approximately two-and-a-half to three-and-a-half hours assessing 300 random test images, DeepCOVID-XR did so in about 18 minutes. It also performed slightly better at 82% accuracy, compared to the radiologists’ accuracy of 76%-81%.
"We feel that this system has the potential to provide significant benefit to overburdened healthcare systems in mitigating unnecessary exposure to the virus by serving as an automated tool to rapidly flag patients with suspicious chest imaging for isolation and further testing," said Wehbe.
The researchers have made the algorithm publicly available so that others can continue training it with new data. DeepCOVID-XR is still in the research phase but may potentially be used in clinical settings in the future.
The entire model and source code used to develop the system are freely available at https://github.com/IVPLatNU/DeepCovidXR.
The findings were published in Radiology