The artificial intelligence (AI) was trained to distinguish cervical conditions that would and would not require treatment.
The image above has been cropped. Image via www.vpnsrus.com
Could cervical cancer be brought under control? Not quite, but the results of a study published in the Journal of the National Cancer Institute seem promising.
Researchers from the National Institutes of Health and Global Good have developed a deep learning algorithm that can analyze digital images of a woman’s cervix and identify precancerous changes that require medical attention — with more accuracy than human experts. The team used comprehensive datasets to train the algorithm to recognize patterns in complex visual inputs, like medical images.
“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 senior author Mark Schiffman, M.D., MPH, of National Cancer Institute’s Division of Cancer Epidemiology and Genetics.
The team created the algorithm by using more than 60,000 cervical images from a National Cancer Institute archive of photos collected during a cervical cancer screening study. The study consisted of more than 9,400 women between 18 and 94 years old and was conducted in Costa Rica from 1993 to 2001. Follow up lasted up to 18 years.
Researchers gained information on which cervical changes became precancers and which did not. The photos were then digitized and used to train the algorithm to distinguish cervical conditions that would and would not require treatment.
This type of artificial intelligence (AI) approach is called automated visual evaluation and according to the authors of the study, has the potential to revolutionize cervical cancer screenings, especially in low-resource settings.
A single visual screening restricted to women aged 25 to 49 years old identified 127 of 228 (55.7 percent) precancers diagnosed in the entire adult population.
A sensitivity of 100 percent was achieved with a specificity of 57.5 percent when the researchers trained another automated visual evaluation model restricted to human papillomavirus (HPV)-positive women.
The algorithm performed better than all standard screening tests at predicting all cases diagnosed during the study and identified precancer with greater accuracy than a human expert review or conventional cytology.
Approximately 80 percent of the estimated 500,000 cervical cancer cases occur in low- and middle-income countries. And HPV and cervical screenings are lacking in lower resource settings. Generally, visual inspection of the cervix after acetic acid application is practical, but is not accurate in distinguishing precancer from common, minor abnormalities.
“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 or Global Good.
The research team will further train the algorithm on a sample of images of cervical precancers and normal cervical tissue from women around the world, using a variety of cameras and imaging options.
Get the best insights in healthcare analytics directly to your inbox.
Read More About AI