The new AI was developed by scientists from Australia's Charles Darwin University. They trained the algorithm using video recordings of lung ultrasounds and concluded that the AI can accurately determine the presence of a problem and even distinguish one respiratory disease from another.
AI identifies specific patterns that indicate different lung diseases. Tests have shown that the diagnostic accuracy is currently 97%: AI was able to successfully differentiate between pneumonia, COVID-19, and other diseases. Unlike humans, AI can see invisible changes in pixels of data that, taken together, are important for the outcome.
It is especially important that in addition to the result, the scientists trained the AI to provide explanations to doctors in order to increase the trust of medical professionals.
In addition to the diagnosis, the AI provides doctors with a report with specific examples in the images from which the corresponding conclusions were made.
The team is currently working on improving the accuracy to make the algorithm universal for diagnosing tuberculosis, cancer, asthma, fibrosis, and many other lung diseases. In the future, such diagnostics can be used at the stage of population screening and when specific diseases are suspected.
One of the common symptoms after COVID-19 is "fog" in the head. Recently, scientists have discovered that the cause of this condition is caused by impaired lung function. Today, such changes in the lungs are diagnosed using MRI, but it is possible that as the new AI learns, it will be able to help with this issue during the rehabilitation stage.