Researchers at the National Cancer Institute developed an algorithm that can accurately identify precancerous changes in the cervix—a potential boon to low-resource areas that don't have enough medical professionals.
In work published Thursday in the Journal of the National Cancer Institute, officials said the NCI team worked with a team from Global Good—a fund at Intellectual Ventures—and "trained" their machine learning algorithm on complex visual inputs, such as medical images. Ultimately, they said, they got the machine to accurately identify patterns of cervical cancer.
They verified the computer's findings with the National Library of Medicine—another branch of the National Institutes of Health—showing the algorithm actually outperformed human experts reviewing the files.
"Our findings show that a deep learning algorithm can use images collected during routine cervical cancer screening to identify precancerous changes that, if left untreated, may develop into cancer," said Mark Schiffman, M.D., M.P.H., of the NCI’s Division of Cancer Epidemiology and Genetics, senior author of the study, in a release. "In fact, the computer analysis of the images was better at identifying precancer than a human expert reviewer of Pap tests under the microscope (cytology)."
The findings are sure to raise further speculation about computers replacing human information workers in the not-too-distant future. But in the present day, when many areas have a severe shortage of healthcare workers, such algorithms could be a big advance for public health.
When low-resource areas do routine cervical screenings today, they're frequently forced to make do with a process called visual inspection with acetic acid, which isn't tremendously accurate. But since it allows healthcare workers to find signs of disease with the naked eye, the process is more convenient and scalable than expensive lab tests.
But with an algorithm like NCI's that allows for automated visual evaluation, healthcare workers can screen for cervical cancer with minimal training. All they need is a camera similar to one on any cell phone, and the screening and treatment can be performed in a single visit.
Cervical cancer is a leading cause of illness and death among women in the developing world.
"When this algorithm is combined with advances in HPV vaccination, emerging HPV detection technologies, and improvements in treatment, it is conceivable that cervical cancer could be brought under control, even in low-resource settings," said Maurizio Vecchione, executive vice president of Global Good, in the release.
NCI's system was trained on 60,000 images taken during a cervical cancer screening study completed in Costa Rica in the 1990s. The 9,400 women who participated in the study were given follow-up lasting for 18 years.
"Because of the prospective nature of the study, the researchers gained nearly complete information on which cervical changes became precancers and which did not," NIH said in the release.
In order to further scale the solution, NCI researchers will need to continue training its algorithm. Cervical cancer can appear differently in various regions of the world, according to the release, so researchers hope to receive representative images from women in communities around the world.
Critically, though, those communities will only be able to use the solution if they are able to access NCI's algorithm. Critics have raised concerns that AI could exacerbate health disparities since low-income areas of the country with less access to digital devices may gradually be left behind.