Publications
Partners
Many women in developing countries lack access to advanced screening for cervical cancer. But researchers say a new "AI" technique might help.
The technique relies on photos and computer artificial intelligence to identify changes that may lead to cervical cancer.
A chance at quality medicine
Catching these changes early, when they're still easily treatable, could help save the lives of many women, said study leader Dr Mark Schiffman and colleagues.
A single round of screening using the new technique could detect dangerous changes in the cervix in more than half of women 25 to 49, the new research found.
"In our first pass [with this technique], we can identify half of women who in the future will get precancer," said Schiffman, of the US National Cancer Institute's division of cancer epidemiology and genetics.
"We want everyone to have a chance at quality medicine and to develop techniques that can reach anywhere to prevent suffering," Schiffman added.
Cervical cancer remains a leading cause of cancer mortality and illness worldwide. A quarter of a million women die from cervical cancer each year, and 90% of these deaths occur in low- and middle-income countries with limited prevention programmes, according to background information in the study.
This type of cancer is caused by a persistent infection with certain strains of the human papillomavirus (HPV). Vaccines are available to prevent HPV.
Convenient and low-cost
In areas where more advanced screening isn't available, health care workers use a technique called visual inspection with acetic acid. A health worker applies diluted acetic acid (a main component of vinegar) to the cervix and then visually examines the cervix. Acetic acid can cause abnormal cells to turn white.
This test is convenient and low-cost. It's also easy to train health care workers to do. But it isn't nearly as accurate as more advanced screening techniques, according to the researchers.
In a study done in Costa Rica in the 1990s, photos were taken of visual inspections of more than 9 400 women. Their health was followed for as long as 18 years. This meant the researchers knew who developed cancer or pre-cancer and who did not.
They used this information and the photos to "train" AI technology to detect who will and won't need treatment.
The computer algorithm was able to detect more suspicious areas (high-grade lesions) and cancers than human experts visually inspecting the cervix, the study found. It even bested the Pap smear, a test that's been the standard for cervical cancer detection for many years in developed countries.
"Machine learning actually outperformed human experts by a considerable margin. It also identified more cancers than a Pap smear. And it has the advantage of requiring very little resources. Eventually, you'll just need a cellphone and you could put an app on it," Schiffman said.
Limited access to medical clinics
Dr Jennifer Wu is an obstetrician-gynaecologist at Lenox Hill Hospital in New York City.
"This new technology will be a tremendous aid to places with low medical resources, where more women die of cervical cancer," said Wu, who had no role in the study.
She said that women in these areas often have limited access to medical clinics. "So visualisation and treatment at the same visit are much better for the patients," Wu noted.
Schiffman said this method probably wouldn't replace methods currently used in the United States. He envisions it being used in areas where women need to be screened and treated on the same day if the tests show any areas of concern.
'See and treat' method
Dr Amanda Shepherd-Littlejohn, an oncologist at NYU Winthrop Hospital in Mineola, New York, agreed.
"Overall, the AI programme is very promising and will be especially useful in areas of the world needing a 'see and treat' method," said Shepherd-Littlejohn, who wasn't involved in the research.
If high-grade lesions or cancer are suspected, the women can have excisions of the abnormal area right then, without waiting for Pap screening or biopsy results, she explained.
Study results were published in the Journal of the National Cancer Institute.
Image credit: iStock
