Algorithms to detect lung lesions: in Brazil, artificial intelligence comes to the aid of doctors to diagnose cases of Covid-19 infection. The RadVid-19 system, developed by the Chinese company Huawei and the German company Siemens, has been operational since April in this country where 2.6 million patients have been identified, of which some 95,000 have died.
Doctors can thus submit an X-ray or CT scan of their patients’ chest for analysis by algorithms to identify spots on the lungs. “The software identifies these areas and estimates the probability of having contracted Covid-19”, explains Marcio Sawamura, deputy director of the Institute of Radiology of the Hospital of Clinics of the Faculty of Medicine of the University of Sao Paulo (USP).
The tool allows you to anticipate
On his computer, the practitioner can analyze the white and yellow circles that identify the lesions. He can also compare the course of the infection with previous images.
Set up with the support, among others, of the Inter-American Development Bank (IDB), the software is now available in 43 Brazilian hospitals, 60% of which are public establishments. Dr Marcio Sawamura stresses that the first assessment established by artificial intelligence does not replace the diagnostic tests prescribed by the doctor.
But the tool allows, according to him, to anticipate the delivery of treatments in this country where the far-right president Jair Bolsonaro has continued to minimize the pandemic and where no massive testing campaign has been put. in place. In five months, Brazil, with a population of 212 million, has performed only 2.7 million PCR virological tests and 4.4 million serological tests.
“Brazil performed fewer tests than it should”
Radiologist Arthur Lobo very frequently uses RadVid-19, “which helps to establish a diagnosis when in doubt.” “Brazil has performed fewer tests than it should, scanners and radios have become diagnostic tools,” explains the practitioner who practices in Belém, the capital of the state of Para (north) where the rate of contamination is 30% higher than the national average.
During the harshest weeks of the health crisis, the system made it possible to anticipate certain diagnoses which took time to be confirmed. “Many times I acted on the clinical history. A patient would come in with symptoms, we would test, but he was negative. Despite this result, I would ask for progressive symptom reviews. X-rays and scanners helped with decision making, ”says Arthur Lobo.
Claudia Leite, professor in the Department of Radiology and Oncology at USP confirms: “Very often we have had patients with alterations in the images, without confirmation of the diagnosis (by test). In these cases, we took the measures intended for Covid-19. Later, we had confirmation that it was indeed Covid-19 “.
Until the end of July, the RadVid-19 device had already made it possible to analyze 10,700 radios or scanners. The reliability of the results is being assessed.
–
