AI-Powered Task-Shifting for High-Quality Fetal Ultrasound Service in Community Healthcare Settings
This project collects high-quality obstetric ultrasound data in Nepal to build and evaluate AI models that support task-shifting. Radiologists at hospitals acquire scans from routine patients to train AI models to detect fetal anatomical views and estimate pregnancy biomarkers. Models are iteratively tested in health posts and hospitals to ensure accurate interpretation by non-specialists.
Many rural and underserved areas in Nepal lack access to specialist ultrasound services. Task-shifting, supported by AI, could allow non-specialist health workers to perform meaningful obstetric assessments safely.
To evaluate how AI can safely enable non-specialists to conduct obstetric ultrasound, including high-risk cases, and to design supervision and training protocols that support task-shifting.
The project has collected diverse blind-sweep scans, trained AI models to detect key fetal anatomical views, and piloted these models in real-world health posts. The initiative supports safe, AI-assisted task-shifting, expanding access to quality obstetric care in underserved regions.