Tested 3,000 women
The study included a total of 3,000 women who would otherwise not have been offered screening for cervical cancer. They visited rural hospitals where cervical cell samples and human papilloma virus (HPV) samples were taken on site, digitised, and analysed using AI. The samples were also examined by pathologists. The researchers trained local nurses, laboratory staff and pathologists to use the system and collaborated with the healthcare authorities to integrate this method into routine healthcare. The women who had signs of cervical cancer subsequently received treatment in accordance with the national guidelines.
AI required consistency in the images
One of the biggest challenges with using AI was that the images it was given to analyse were not always sufficiently consistent. To make cells visible in the samples under a microscope, the cells are stained. Staining reagents and thus cell colour could differ between countries and deliveries, which meant that the images that the AI was asked to analyse were not always consistent enough.
“The AI method worked well technically, but unreliability in the supply of reagents, variations in reagent quality and power outages all affected accuracy as well as the capacity to perform these tests rapidly, including HPV analyses,” says Nina Linder.
Another difficulty was finding the women who had shown signs of cancer and needed follow-up care.
“In Tanzania, we had quite a few problems with follow-up. Some women did not come back, and when we later checked their samples, it turned out that they had changes that needed treatment. Sometimes it is difficult for local doctors to find the patients and get them to understand that they need treatment. We followed up as best we can and tried to give all women the opportunity for further investigations,” says Nina Linder
Can increase trust in the healthcare system
Although the study describes both opportunities and challenges with the AI method, the researchers see it as a first step in evaluating AI-supported diagnostics in more comprehensive healthcare programmes and for more women’s diseases.
“For decades, diagnostic methods that are proven to be effective for women’s health – such as cell-sample based cervical cancer screening – have been dependent on highly trained experts. With the latest advances in medical AI, we can now re-evaluate these methods and introduce them even in resource-limited settings, making life-saving diagnostics far more accessible,” says Johan Lundin, professor at Karolinska Institutet and one of the co-authors of the study.
Another valuable contribution is that it raises awareness locally of why screening is important.
“When women see that there is reliable healthcare to go to and that they do get help, it lowers the threshold to seek care, which strengthens health as well as social engagement,” says Nina Linder.
Reference: Linder N, Nyirenda D, Mårtensson A, Kaingu H, Ngasala B, Lundin J. AI supported diagnostic innovations for impact in global women’s health. BMJ. 2025;391:e086009. doi: 10.1136/bmj-2025-086009
This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.
