Building Technology and Trust: AI Algorithms and Cervical Cancer
- dukegwht
- 1 hour ago
- 3 min read
In Kisumu, a city in western Kenya, a team of researchers and clinicians is working to make cervical cancer screening more accessible and more human. The Duke Center for Global Women's Health Technologies (GWHT), in collaboration with The Center for Global Reproductive Health, and the Kenya Medical Research Institute (KEMRI), are creating sustainable and affordable models for early detection of this highly preventable disease.
But building effective health solutions require more than technology and algorithms: it demands deep partnership with the communities and providers who will ultimately use them.
Bringing Care Closer to Women
For many women in western Kenya, healthcare can feel out of reach. Clinics are far apart, specialists are scarce, and basic equipment is limited. These barriers mean that conditions like cervical cancer are often caught too late.
This collaboration aims to change that by introducing self-collection swabs that women can use to obtain cells that can be tested in a lab for HPV, the virus that causes cervical cancer, allowing women to take the first step themselves. By equipping local clinics with mobile tools that make patient and health care worker communication and scheduling of follow up care reliable.
At the clinic, providers use a handheld imaging device called the Pocket Colposcope to take detailed photos of the cervix. The images are uploaded through a mobile app, the Calla Health App (CHA), which also helps train an AI algorithm to assist with diagnosis and follow-up. The Pocket Colposcope closes an important clinical gap, triaging patients who test positive for HPV.
Smarter Tools, Not Replacements
Artificial intelligence has the potential to level the playing field in global health. In places where trained specialists are few, AI can help ensure more consistent and accurate results.
But, as the team quickly learned, building good AI isn’t just a technical challenge: it’s a human one.
Image quality varies, internet connections drop, and every provider has a slightly different way of working. One common issue is blur in handheld images, which can reduce accuracy. To fix this, the team built an AI software tool that detects when an image is too blurry and alerts the provider to retake it. It also helps developers to evaluate and take into account image quality when they create and test algorithms.
The idea worked in theory, but in practice, it sometimes slowed down clinic workflow. For the researchers, that was an important reminder: real-world solutions need real-world input.
“We’re not just looking at the images,” says Dr. Erica Skerrett, one of the project’s lead researchers. “By talking to providers, we’re learning about the challenges they face day to day and how technology can fit better into their workflow, not disrupt it.”
Building Technology That Listens
Effective AI depends on more than good data. It requires understanding the context. By maintaining an ongoing dialogue with healthcare workers, the team is building not just better algorithms, but mutual trust. That trust helps ensure that when technology doesn’t work as expected, providers can speak up and developers can adapt.
This work in Kenya offers a model for what responsible health innovation can look like: community-driven, transparent, and grounded in empathy.
Instead of building technology first and adapting it later, the GWHT team is designing with and not just for the people who will use it. The result is more than a diagnostic tool. It’s a partnership built on respect, communication, and shared commitment to saving lives.
