Artificial intelligence could be much more accurate than humans and traditional tests to detect cervical pre-cancers, according to US researchers Thursday.
Cervical cancer is the fourth most common cancer in the world, with 570,000 cases per year in 2018 according to the World Health Organization (WHO). In rich countries, advances in screening and vaccination campaigns against HPV (linked to these cancers) have reduced mortality, but these advances benefit less poor countries, which accounted for 90% of deaths in 2012, according to WHO .
“Cervical cancer is now a disease of poverty,” says lead author of the new study, Mark Schiffman, a doctor at the National Cancer Institute near Washington, who has been working for 35 years against this cancer. “We are trying to find extremely affordable and easy, but very accurate methods to attack cancer by the vaccine or by simple techniques, such as a smartphone,” he says.
The doctor is part of a team that has developed an algorithm from an archive of 60,000 images of cervical uteri taken in Costa Rica in the 1990s.
A total of 9,400 women participated and were followed for several years, for 18 years for some – which made it possible to link each cancer that appeared among them to the image of the cervix before the onset of the disease.
Once trained on this image basis, the algorithm visually detected precancerous cells in 91% of cases, according to the study published Thursday in the journal of the National Cancer Institute.
You should also read: