AI algorithm helps Catharina Hospital screen SEH patients for COVID-19

A new algorithm developed by Catharina Hospital in Eindhoven speeds up the screening of patients in the Emergency Department (ED) on COVID-19. The algorithm uses artificial intelligence (AI) to recognise the effects of corona in the blood.

Patients arriving in A&E are routinely screened for COVID-19. This blood is then tested for 30 different values. However, the peak of the coronapause put this testing capacity under pressure. The AI algorithm will help better handle such a peak in the future.

Clinical chemist Dr Arjen-Kars Boer: "At the height of the coronapause, we were asked if there wasn't a faster test to rule out COVID-19. In the Emergency Department, it is important to know as soon as possible whether a patient is coronapositive or not. To do this, we started looking at the information we already had. In fact, patients who come into the ER have their blood taken as standard for a so-called quick scan. Their blood is tested for as many as 30 different values. We wondered whether their blood sometimes showed subtle changes characteristic of COVID-19. Couldn't we find some kind of instantly recognisable barcode? And yes, so it turns out we could! But we didn't get it as a gift..."

For the development of the algorithm, data from over 10,000 patients at the A&E were converted into digital data. This data was then analysed using AI. "Then we had to look for the smallest set of results that make a good prediction," says bio-statistician ir Ruben Deneer. Deneer interrupted his TU/e PhD research for a three-month period to use algorithms to analyse the dataset. "Eventually, in a very short time, we managed to identify ten laboratory tests that together constitute the most important predictors of COVID-19," says the bio-statistician.

Boer: "At Catharina Hospital, we are very good at big data and we therefore see it as our social responsibility to tackle this kind of challenge. We did it alongside our regular work, which was not a sine cure, but we need to improve tomorrow's care today. Developing a tool like this in three months is really unique. In the whole world, this has not yet been described in the literature."

new tool is now in use at Catharina Hospital by A&E doctors. "The big advantage of this test is that we can now handle care much more efficiently. Now we test, so to speak, 100 people with a PCR test, the standard corona test so to speak, when it only needed to be 10. At the same time, with this new tool, we see corona symptoms in the blood of people who, because they have no recognisable symptoms, are not tested for corona. We can now offer them a PCR test after all, because if there is something we don't want is for infected people to end up in a 'regular' hospital ward," Boer says. He emphasises that the algorithm is not a replacement for the PCR test, but rather an additional tool for doctors.

In time, Boer and Deneer hope to deploy the algorithm in other hospitals as well. "That still has some way to go, because each hospital performs different tests in the ER, but eventually it is only a matter of time that other hospitals will adopt our tool."

Other hospitals in the Netherlands are also working on deploying AA in the fight against COVID-19. For example, the Jeroen Bosch Hospital recently reported seeing several possibilities for the deployment of AI during the coronacrisis, based on research. Among other things, to address the scarcity of test kits to test for COVID-19.

"Radiography is a fast and relatively cheap imaging modality available in many healthcare settings where resources are scarce. Unfortunately, there is a severe shortage of radiological expertise in this region for accurate interpretation of such images. An AI system could be a useful tool to complement radiologists or in the common case where radiological expertise is not available," write researchers Keelin Murphy, Henk Smits, Arnoud J.G. Knoops, Mike B.J.M. Korst, Tijs Samson, Ernst T. Scholten, Steven Schalekamp, Cornelia M. Schaefer-Prokop, Rick H. H. M. Philipsen, Annet Meijers, Jaime Melendez, Bram van Ginneken and Matthieu Rutten in an article in the scientific journal Radiology.

AI software has previously been experimented with to detect tuberculosis on medical images. The new system tested by the Jeroen Bosch Hospital builds on this and is called CAD4COVID-XRay. The system was developed by Delft Imaging and Thirona. The AI software analyses chest X-ray images of patients potentially infected with COVID-19. The software rates these images with a score from 0 to 100, reveals COVID-19-related abnormalities in the lungs and shows which part of the lungs is affected.

AI system uses deep learning, a form of AI where an algorithm is trained by having it search for patterns in a large dataset. A characteristic of a deep learning algorithm is its ability to keep learning and getting smarter.

Jeroen Bosch Hospital trained this system using a dataset of 22,184 lung images, of which 7,851 were of healthy lungs and 5,012 of pneumonia. Here, a dataset of lung photos of general pneumonia was deliberately chosen, as a large dataset of lung photos with pneumonia caused by COVID-19 is difficult to obtain. The algorithm was further trained after this initial training to specifically identify COVID-19.

algorithm was then tested, with 454 hospital X-ray images reviewed by both the AI system and six radiologists. During the experiment, the algorithm was found to be able to assess all images as well as radiologists. More information is available here.

Author: Wouter Hoeffnagel