Equitable AI-Powered Cervical Cancer Screening in Nepal
Summary
In 2024, Dhulikhel Hospital – Kathmandu University (DH-KUH) committed to scale cervical cancer screening and treatment in Nepal using its AI-powered clinical decision support app that captures cervical images, provides real-time clinical decisions, and supports patient follow-up. DH-KUH aims to reach 302,000 women over the next five years, doubling the country’s cancer screening rates. DH-KUH will use initial screening data to increase accuracy of the app and will train 1,050 frontline healthcare providers and 6,600 community health volunteers. These trained cadres will then conduct cervical cancer screening camps in partnership with local governments, targeting communities with limited access to screening. Providers will provide prompt treatment at the health facility and refer to tertiary hospitals for those who need additional care. By using AI to expand access to cervical cancer screening and enabling earlier detection and treatment of cervical cancer, DH-KUH and partners will save women’s lives.
Approach
Dhulikhel Hospital – Kathmandu University (DH-KUH) commits to scaling cervical cancer screening and treatment in Nepal using its AI-powered clinical decision support app that captures cervix images, provides real-time clinical decisions, and supports patient follow-up. DH-KUH aims to reach 302,000 women over the next five years, doubling the country’s cancer screening rates.
Firstly, DH-KUH will refine the app to enhance its sensitivity (currently at 74%k;op-0) and specificity (currently at 76%) to 85% using data from the first 42,000 screenings. Upgrades will also include patient registration, customized forms tailored for specific needs, and offline capabilities. Scale of the AI-powered cervical cancer screening will then include building local health system capacity and extending reach to women in rural and remote communities. Initially, DH-KUH will train 1,050 frontline healthcare providers and 6,600 community health volunteers. These trained cadres will then conduct cervical cancer screening camps in partnership with local governments, targeting communities with limited access to screening. Providers will provide prompt treatment at the health facility and refer to tertiary hospitals for those who need additional care. To ensure consistent data collection, the app has a built-in reminder for the service provider that will buzz every 15 minutes until the final treatment result is entered.
The DH-KUH team has extensive experience and expertise in delivering community-based services across various health sectors. Their multidisciplinary team includes epidemiologists, gynecologists, implementation scientists, IT and health system experts, and field workers who will facilitate the project. Additionally, they will leverage their long-standing partnership with AMPATH-Global, local artificial intelligence expert organizations, and the Nepal government to support the initiative. Training in proficient use of the app will ensure the AI-powered clinical decision support app is integrated into the existing healthcare infrastructure. Consistent technology support to healthcare workers will also ensure data quality and security.
Action Plan
Year 1 (October 2024 2025 to December 2025) :
Q1, the team will initiate project setup and planning.
Q2, finalize protocols, SOPs, and the training manual, with comprehensive staff training.
Q3 to Q4, begin the first phase of data collection, validating the application with 16,000 women screened and 112,000 expert-annotated images.
Year 2 (January 2026 to December 2026) :
The team will screen a total of 26,000 women.
Q1 to Q3, screen a total of 8000 women each quarter
Q4, screen a total of 2000 women. Finalize the deployment system and develop training materials to support the ongoing implementation and scaling efforts in Q4.
Year 3 (January 2027 to December 2027) :
The team will train 1050 health workers to use the AI-VIA app and community health volunteers to conduct massive awareness in the community. Training will run concurrently with screening efforts and treatment efforts to reach 130,000 women.
Q1 to Q4, screen approximately 32,500 and treat 2,600 each quarter.
Year 4 (January 2028 to December 2028) :
The team will continue its effort to screen 130,000 women, screening 32,500 women and treating 2,600 positive cases per quarter.
Q1 to Q4, screen approximately 32,500 and treat 2,600 each quarter.
Year 5 (January 2029 to December 2029) :
The team will continue its effort to screen 130,000 women, screening 32,500 women and treating 2,600 positive cases per quarter. The team will disseminate the project results to national and international stakeholders.
Q1 to Q4, screen approximately 32,500 and treat 2,600 each quarter.
Background
Cervical cancer (CC) , the second leading cause of cancer-related deaths in women worldwide and disproportionately impacts low and middle-income countries (LMICs) , where 90% of global cervical cancer deaths occur (WHO 2024) . In Nepal, CC is the most common cancer among women, with approximately 2942 cases and 1928 deaths occurring annually.(GBD 2019 Risk Factors Collaborators, 2020) .
While cytology-based screening programs have drastically reduced cervical cancer deaths in high-income countries, their implementation requires complex infrastructure and high-quality healthcare services, which most low-income countries cannot afford. In Nepal, the CC screening coverage remains low at only 8% (Nepal STEPS Survey 2019) . The Government of Nepal has prioritized an equally effective method using the Visual Inspection Acetic Acid (VIA) approach for screening, which is safe, inexpensive, and easily administered by mid-level health providers, with immediate result interpretation, enabling prompt treatment in a single visit (MOHP 2010) .
However, VIA test sensitivity varies, ranging from 41% to 82% across settings, primarily due to healthcare providers’ skills (Katanga 2019; Gravitt PE 2010; Ardahan M 2011; Campos NG 2015) . VIA relies on subjective visual assessment, leading to inconsistent results, with physicians achieving better outcomes than nurses and paramedics. Additional barriers to screening and treatment are a lack of trust in health providers’ competencies, an unprepared healthcare system, and mountainous terrain that hinders access (Raifu et al. 2017) .
In response to these challenges, Dhulikhel Hospital-Kathmandu University Hospital (DH-KUH) developed an AI-powered VIA decision support system. A mobile app allows healthcare providers to capture and submit cervix images, and an AI model processes the images, provides real-time positive or negative clinical decisions, and supports patient follow-up of test-positives. This AI-powered VIA decision support system reduces health worker burden, improves clinical accuracy, and increases trust in the health system, leading to scaled access and treatment in rural and remote areas
Progress Update
Partnership Opportunities
Dhulikhel Hospital is seeking additional financial resources to build upon the existing work. This will allow them to deploy a validated AI-powered cervical cancer clinical decision App and reach 30,200 women for screening and treatment of cervical cancer, a preventable disease., AMPATH Nepal has partnered with Dhulikhel Hospital-Kathmandu University Hospital to build holistic and sustainable health solutions in Nepal, and together, they are committed to improving health outcomes by leveraging their combined expertise and resources. DH-KU is strongly committed to ethical considerations and regulatory compliance to ensure projects are conducted with integrity and accountability. This partnership provides access to a highly skilled team of health professionals with extensive experience in cervical cancer screening and diagnosis, as well as project management and implementation. Through its extensive medical infrastructure, DH will provide expert human resources, time (in kind) , physical spaces, and necessary equipment for training, implementing, and monitoring the project.