Diagnostic App for Cervical Cancer

In yet another example of machine learning, researchers have developed an algorithm to screen women for cervical cancer using a digital photograph. Known as “automated visual evaluation,” this artificial intelligence (AI) technique could revolutionize screening methods, particularly in low-resource environments.

The most common way to identify cancerous or precancerous cells in the cervix is through a pap smear in which cell samples are sent to a lab for analysis. But in regions that lack the needed laboratory resources, doctors use a method that involves swabbing the cervix with dilute acetic acid and visually inspecting it for white spots that could be indicative of cancer. Instead of a simple visual inspection, researchers decided to photograph more than 9400 cases and use the images to train a deep learning algorithm that detects abnormal tissue. The results were then compared to findings by independent experts and it was concluded that the computer performed better than human reviewers regardless of whether the testing was done using the acetic acid method or a pap smear.

The developers plan to further refine the algorithm and adapt it for use on mobile phones.