Curable if detected early enough and appropriately treated, cervical cancer is the 4th most common cancer among women globally and the leading cause of cancer deaths for women in sub-Saharan Africa. In 2020, around 342,000 women died of cervical cancer worldwide – the WHO projects this number to reach 416,000 by 2035. More than 85% of those deaths occur in low- and middle-income countries.

An initative to find solutions to this context was launched by Prof. Patrick Petignat of the Hôpitaux Universitaires Genève (HUG) in partnership with Prof. Jean-Philippe Thiran, who leads the Signal Processing Laboratory 5 of the Swiss Federal Institute of Technology (EPFL). The project partnership includes Bafoussam and Dschang Regional Hospitals in Cameroon. SmartCervix is Integrated into initiative and coordinated at EssentialTech.

Treatment of precancerous and cancerous lesions, with cryotherapy or thermal ablation for instance, can prevent cervical cancer from developing. If cervical cancer is diagnosed early enough, survival increases significantly the sooner it is treated.

Unfortunately, cervical cancer screening is not yet routine for most women in low- and middle-income countries and in “medical deserts” in general. On a technical level, there is a severe lack of infrastructure and equipment in many regions of the world. At the social level, there is a lack of information, awareness and education about women's reproductive health in the community. Finally, at the economic level, there is a lack of trained staff as well as a lack of financial resources for women to be able to access the screening services.

Our adapted solution: An AI-based clinical decision support tool for cervical cancer screening

Technical aspect

The smartCervix project, coordinated at EssentialTech, aims to improve cervical cancer screening. It includes an automated artificial intelligence-based image analysis solution to reduce subjectivity and risk of misinterpretation of visual inspection:

•  Image sequences of the cervix are captured with a smartphone

• They are analyzed using artificial intelligence

• The algorithm evaluates aceto-whitening, which can indicate the presence of cervical lesions

• The images are interpreted by a specialist locally or by telemedicine

Socio-economic aspect

A tool alone cannot make change.

The smartCervix project also includes a socio-economic component, including research into the most appropriate deployment model in these vulnerable contexts. This is aligned with our overall research strategy at the EPFL EssentialTech Centre, which aims to ensure that technology is used within the local systems, is financially affordable and is maintained over the years. Applying this strategy is the only way to ensure that women in vulnerable contexts have stable access to cervical cancer prevention and screening.

Service delivery is specifically important to this project. To ensure that women attend the screening program regularly, community health workers who raise awareness of cervical cancer among women in villages, churches and women's associations to encourage screening are critical. Local nurses or midwives are trained to use the smartphone to collect images during the VIA/VILI procedure.

The project partnership also includes the Bafoussam Regional Hospital and the Dschang District Hospital in Cameroon.

Completed steps:

•  We have developed a mobile app to acquire the images necessary to train the algorithm for the AI to function. 

•  A data collection campaign (using the same application) has generated data for over 2000 patients in Switzerland and Cameroon

• Preliminary results show that technology outperforms the health care practitioners (colposcopists)

Current activity:

• The algorithm is currently being consolidated and will soon be ready for deployment

• A deployment model for Cameroon and other vulnerable health systems is being developed to ensure that the solution has a long-term impact