CERVICAL cancer is the leading cause of cancer-related deaths among women in Zimbabwe, claiming 1,796 lives annually. This equates to a full Honda Fit — the driver, a front-seat passenger, and four back-seat passengers — perishing in a car crash every single day. These grim statistics, however, are largely preventable.
Cervical cancer, caused by the Human Papilloma Virus (HPV), develops slowly — giving women a ten-year window for early detection through screening.
The Government of Zimbabwe has made notable strides in combating the disease. This includes training nurses in the Visual Inspection with Acetic Acid and Cervicography (VIAC) method, along with the traditional Pap smear.
In 2011, Zimbabwe introduced VIAC as a free service in public hospitals as part of a national initiative to enhance reproductive health and fight cervical cancer. The aim was to improve access to screening for this deadly disease.
However, Zimbabwe continues to experience a persistent brain drain, with trained nurses and healthcare workers often leaving for better opportunities abroad.
VIAC, while suitable for low-resource settings, has a sensitivity rate of only 60%. This means that 40% of women with cervical cancer might receive a false negative — being told they are cancer-free when they are not.
Realizing this critical gap, Dr. Lindelwe Ncube, a medical doctor and biomedical engineer, saw the potential of artificial intelligence (AI) to bridge it. He developed a tool dubbed ColpoPen, which leverages AI to enhance cervical cancer screening.
“AI is often demonized as a technology that takes away jobs. But in low-resource settings plagued by brain drain, it can be exactly what the doctor ordered,” said Dr. Ncube.
Explaining the VIAC process, he said, “Nurses apply acetic acid to the cervix and observe for any whitish areas, which may indicate pre-cancerous lesions. In Zimbabwe, we follow a ‘screen-and-treat’ model where these lesions are removed immediately.”
“But VIAC only gives a true positive result 60% of the time. That means in 40% of cases, it falsely indicates there’s no cancer when there actually is,” he explained.
To address this, Dr. Ncube and his team began researching the integration of AI to boost accuracy. Using annotated images from VIAC and colposcopy procedures — identified by experts as either showing cervical cancer or not — they trained, validated, and tested an AI model.
They found that the model had a higher sensitivity rate than VIAC alone. As a result, they incorporated it into a mobile application linked to a magnifying camera, enabling real-time cervical analysis.
“With this app, the nurse can examine the cervix while the AI simultaneously analyzes the image. The goal is to catch the 40% of cases VIAC might miss, potentially raising sensitivity to around 90%,” said Dr Ncube.
The app also includes a telemedicine feature, allowing nurses to consult with doctors or specialists during examinations. “The nurse, the specialist, and the AI model work together — increasing diagnostic accuracy and improving patient outcomes,” he added.
A major advantage of the app is the creation of a national cervical image database, which can be used for further research and AI development. “It’s a homegrown solution,” said Dr. Ncube. “Instead of purchasing external datasets, local researchers can use our own images to improve Zimbabwe’s healthcare system.”
Dr Ncube credits Education 5.0 — a government initiative promoting innovation and industrialization in higher education — for supporting the development of ColpoPen.
At Zimbabwean universities, students are now required to identify real societal problems and develop practical solutions.
“When we were studying, we were told that you had to create something or you wouldn’t graduate. You even had to spin off a company from your idea,” he said. “That’s how ColpoPen was born — a local solution to a national problem.”
As they continue collecting more images, Dr Ncube’s team is working to refine the model’s accuracy. “The cervix differs depending on age and childbirth history.
For example, a woman who has never given birth has a different cervical anatomy compared to one who has delivered three children,” he explained.
“So, if the model was trained only on certain types of images, it may be less accurate for others.”
He emphasized the need for broader data collection to minimize such bias. “Once the model is validated across different age groups and anatomical variations, and passes specialist review, we can confidently roll it out nationally,” he said.
By combining local innovation with modern technology, Zimbabwe is taking a bold step toward reducing cervical cancer deaths — and AI may just be the game-changer the country needs.
