AI program provides high number correct Covid-19 diagnoses in low-income countries Interesting alternative for African regions with shortage of medics

In regions where high-tech laboratory tests are not available, proper Covid-19 diagnoses can be established by an AI program, Radboudumc researchers with international colleagues write in Scientific Reports. The AI program uses an X-ray and one or two blood tests that can be read on the spot. A customized version of the program may eventually be used to diagnose tuberculosis as well.

Covid is diagnosed in the Netherlands and many other countries through a cotton swab in the nose, throat or both. A PCR test is then used to analyze whether fragments of the virus are in the cotton wool. Yes? Covid. No? - no Covid. Problem is the high-tech PCR (polymerase chain reaction) test is not available everywhere in the world. Then other techniques are needed to diagnose infection with the virus.

X-ray

In Scientific Reports, Keelin Murphy and Bram van Ginneken of Radboudumc and international colleagues describe an AI program that could be a good alternative to the rapid antigen test. The program combines an X-ray of the lungs with Point Of Care (POC) blood tests. The diagnostic value of the X-ray by itself is too low, because the majority of infected people have only mild symptoms and no pneumonia is detectable either. But combined with the results of one or more blood tests, it becomes a reliable test.

Point Of Care

Point Of Care tests show directly the results of what you want to measure so you don't need a laboratory. One POC test (white blood cell counts WBC) counts the number of different white blood cells. Higher counts indicate an infection or inflammation. The other POC test measures C-reactive protein (CRP). If it has a higher than normal value, it’s also an indication of inflammation. When the X-ray data are combined with WBC, eventually supplemented by CRP, a diagnosis emerges that’s more sensitive than an antigen rapid test.

Four locations in Africa

The tests took place in four different locations in Africa where there is a lack of people and resources to make Covid diagnoses using PCR tests. In those places, good X-rays of the lungs can be taken, which can also be checked for tuberculosis immediately with software such as CAD4TB, as Thirona and Delft Imaging show. The POC tests fit as well. In total, over 3,500 people participated in the study and a PCR test was performed for verification. Not only could the accessible X-ray and blood tests be performed well on site, but compared with the rapid antigen test, the X-ray-blood test combination, assessed by an AI program, was more sensitive than the antigen test. More sensitive means the test identifies more correct positive diagnoses.

Application with perspective

"This study is the first to validate AI tools for COVID-19 detection in an African setting", the researchers write. "It demonstrates that screening for COVID-19 using AI with point-of-care blood tests is feasible and can operate at a higher sensitivity level than antigen testing.” Looking further ahead, they note that AI programs that combine X-ray with POC blood tests may eventually be suitable for diagnoses of tuberculosis in medically underserved regions.

­­­­-

Paper in Scientific Reports: COVID-19 screening in low resource settings using artificial intelligence for chest radiographs and point-of-care blood tests - Keelin Murphy, Josephine Muhairwe, Steven Schalekamp, Bram van Ginneken, Irene Ayakaka, Kamele Mashaete, Bulemba Katende, Alastair van Heerden, Shannon Bosman, Thandanani Madonsela, Lucia Gonzalez Fernandez, Aita Signorell, Moniek Bresser, Klaus Reither & Tracy R. Glass

• Want to know more about these subjects? Click on the buttons below for more news.

  News homepage news & mediaArtificial intelligence