Finnish scientists collaborating with a small village clinic in Kenya have developed an Artificial Intelligence (AI) program that can quickly diagnose cancerous cells.
In the ongoing study, reported last week in Nature Medicine the team uses a small handheld microscope to sends images from cervical samples to the cloud for analysis.
The researchers from the University of Helsinki, Finland, working with Kinondo Health Center at Kenya’s Coast are reporting significant successes.
Reporting on the study, science journalist Nicole Wetsman of New York, US, says once images are uploaded they are processed with an artificial intelligence algorithm that can rapidly identify abnormal cells.
“We don’t have enough pathologists in this part of Kenya to take care of people. This will reduce the turnaround time,” Harrison Kaingu, head of the Kinondo clinic had told Wetsman
So far, she explained, the team has collected Pap smears from 720 patients at the health center.
Around half of those samples were used to train the algorithm, and the remainder to confirm that the as-yet-unpublished algorithm worked.
Right now, Johan Lundin, the project leader says, the algorithm has a sensitivity of around 95 percent and a specificity of around 85 percent. “The next step is to implement what we have developed.”
However, the author explains that the analysis will still be done using a combination of human analysis and artificial intelligence.
“Right now, it’s really well suited for screening and preprocessing samples. It pulls out areas that look suspicious, and the pathologist only has to look at those spots to confirm,” said Lundin.