New AI System Rapidly and Accurately Detects COVID-19 Virus

Scientists from Northwestern University have designed an advanced artificial intelligence (AI) platform that can spot the COVID-19 virus by studying X-ray images of the lungs.

Scientists from Northwestern University have designed an advanced artificial intelligence (AI) platform that can spot the COVID-19 virus by studying X-ray images of the lungs.
Aggelos Katsaggelos. Image Credit: Northwestern University.

The machine-learning algorithm, known as DeepCOVID-XR, detected the COVID-19 virus in X-rays at around 10 times faster and 1% to 6% more precisely when compared to a group of specialized thoracic radiologists.

According to the researchers, the AI system could be used by physicians to quickly screen even those patients who are brought to hospitals for reasons other than the COVID-19 infection.

Faster and earlier detection of the highly infectious virus could protect patients and health care workers by allowing the hospitals to isolate the positive patient sooner. The authors of the study further believe that the new algorithm could also flag patients for isolation and testing purposes who are not otherwise being evaluated for the COVID-19 disease.

The study was recently published in the Radiology journal on November 24th, 2020.

We are not aiming to replace actual testing. X-rays are routine, safe and inexpensive. It would take seconds for our system to screen a patient and determine if that patient needs to be isolated.

Aggelos Katsaggelos, Study Senior Author and AI Expert, Northwestern University

Dr. Ramsey Wehbe, a cardiologist and postdoctoral fellow in AI from the Bluhm Cardiovascular Institute at Northwestern Medicine, stated, “It could take hours or days to receive results from a COVID-19 test. A.I. doesn’t confirm whether or not someone has the virus. But if we can flag a patient with this algorithm, we could speed up triage before the test results come back.”

Katsaggelos is the Joseph Cummings Professor of Electrical and Computer Engineering in McCormick School of Engineering at Northwestern University. He also has courtesy appointments in radiology and computer science. Wehbe is a postdoctoral fellow from the Bluhm Cardiovascular Institute at Northwestern Memorial Hospital.

A Trained Eye

For a majority of the patients with COVID-19 infection, chest X-rays reveal analogous patterns; for example, the lungs of these patients appear hazy and patchy instead of clear and healthy organs.

Many patients with COVID-19 have characteristic findings on their chest images. These include ‘bilateral consolidations.’ The lungs are filled with fluid and inflamed, particularly along the lower lobes and periphery.

Dr Ramsey Wehbe, Cardiologist and Postdoctoral Fellow in AI, Bluhm Cardiovascular Institute, Northwestern Medicine

The main issue is that heart failure, pneumonia, and other diseases in the lungs can appear almost the same on X-rays. Only a trained eye can spot the difference between a COVID-19 infection and something less infectious.

The laboratory of Katsaggelos focuses on applying AI platform for medical imaging. Both Katsaggelos and Wehbe had already been working closely on cardiology imaging projects and speculated whether they could design a novel system to help combat the COVID-19 pandemic.

When the pandemic started to ramp up in Chicago, we asked each other if there was anything we could do. We were working on medical imaging projects using cardiac echo and nuclear imaging. We felt like we could pivot and apply our joint expertise to help in the fight against COVID-19.

Dr Ramsey Wehbe, Cardiologist and Postdoctoral Fellow in AI, Bluhm Cardiovascular Institute, Northwestern Medicine

AI Versus Human

To design, train, and test the novel AI algorithm, the team used as many as 17,002 chest X-ray images. These images were the biggest published clinical dataset of chest X-rays collected from the COVID-19 period and were utilized for training an AI system.

Among these images, 5,445 came from patients infected by COVID-19 and residing at locations across the Northwestern Memorial Healthcare System.

The researchers subsequently tested the DeepCOVID-XR system against five specialized cardiothoracic fellowship-trained radiologists on 300 haphazard test images obtained from Lake Forest Hospital. All radiologists took around two-and-a-half to three-and-a-half hours to study the set of images, while the AI system took just around 18 minutes.

The accuracy of the radiologists ranged between 76% and 81%, whereas the DeepCOVID-XR system performed somewhat better at an accuracy of 82%.

These are experts who are sub-specialty trained in reading chest imaging,” added Wehbe. “Whereas the majority of chest X-rays are read by general radiologists or initially interpreted by non-radiologists, such as the treating clinician. A lot of times decisions are made based off that initial interpretation.”

Katsaggelos added, “Radiologists are expensive and not always available. X-rays are inexpensive and already a common element of routine care. This could potentially save money and time—especially because timing is so critical when working with COVID-19.”

Limits to Diagnosis

Evidently, not all COVID-19-infected patients display any sign of illness, including the one on their chest X-rays. Particularly during the early progression of the virus, patients are not likely to have manifestations on their lungs.

In those cases, the A.I. system will not flag the patient as positive. But neither would a radiologist. Clearly there is a limit to RADIOLOGIC diagnosis of COVID-19, which is why we wouldn’t use this to replace testing,” concluded Wehbe.

The researchers from Northwestern University have made the new algorithm publicly available so that other scientists can continue to train it with the latest data. At present, the DeepCOVID-XR system is still in the research stage, but it could perhaps be applied to clinical settings in the days to come.

The coauthors of the study include Jiayue Sheng, Shinjan Dutta, Siyuan Chai, Amil Dravid, Semih Barutcu and Yunan Wu, all members of Katsaggelos’ laboratory—and Drs Donald Cantrell, Nicholas Xiao, Bradly Allen, Gregory MacNealy, Hatice Savas, Rishi Agrawal, and Nishant Parekh, all radiologists from Northwestern Medicine.

Journal Reference:

Wehbe, R. M., et al. (2020) DeepCOVID-XR: An Artificial Intelligence Algorithm to Detect COVID-19 on Chest Radiographs Trained and Tested on a Large US Clinical Dataset.
Radiology. doi.org/10.1148/radiol.2020203511.

Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type
Submit

While we only use edited and approved content for Azthena answers, it may on occasions provide incorrect responses. Please confirm any data provided with the related suppliers or authors. We do not provide medical advice, if you search for medical information you must always consult a medical professional before acting on any information provided.

Your questions, but not your email details will be shared with OpenAI and retained for 30 days in accordance with their privacy principles.

Please do not ask questions that use sensitive or confidential information.

Read the full Terms & Conditions.