Scientists at Beth Jezreiel Medical Center in Boston have created artificial intelligence microscopes that can themselves look for dangerous bacteria in a blood sample. The results are published in the Journal of Clinical Microbiology. This is reported by Eurekalert.
With a shortage of highly qualified microbiologists, it is difficult to diagnose patients. But scientists have found a solution to this problem. They took an automatic microscope that can take high-resolution photographs and blood samples that were stained to make the bacteria better visible. They then used an ultra-precise neutron network designed for image recognition and trained it to recognize bacteria by their shape and location.
The researchers started with the most common bacteria: E. coli with its elongated shape, staphylococcus, whose round cells grow in "clusters", and streptococcus, which can grow in pairs or chains. Scientists have made more than 100,000 pictures for teaching out of 25,000 images obtained from everyday clinical practice. Artificial intelligence was able to recognize three species of bacteria with 95% accuracy. The researchers then proposed that 189 samples be sorted without any human intervention. The algorithm coped with the task and gave correct answers 93% of the time.
“As a child, this system needs training,” said lead author James Kirby, director of the Clinical Microbiology Laboratory and professor at Harvard Medical School. "Learning takes a lot of practice, the system makes a lot of mistakes and learns from them."
A quick diagnosis of bacteria in the blood can save the lives of many people: now a patient with blood poisoning has a 40% risk of death. According to the authors of the study, artificial intelligence can help not only determine the causes of infection in medical centers where there are not enough microbiologists, but can also educate students and trainees.