PROJECT TEAM
Professor Martin G. Tolsgaard (Rigshospitalet). Role: Medical lead Professor
Mads Nielsen (Dept. Computer Science, University of Copenhagen). Role: Commercial lead
Professor Aasa Feragen (DTU). Role: Deep learning specialist Associate
Professor Anders Nymark Christensen. Role: Deep learning specialist + QMS
Associate Professor Morten Bo Svendsen. Role: Systems integration + QMS
THE NEED
In many countries, including Denmark, routine scans are limited to the first and second trimester. Without a third-trimester growth scan, only 19–32% of growth-restricted fetuses are detected. Even when performed, scan accuracy varies widely due to operator skill and manual measurement errors. Current methods rely on selecting a few images and using basic formulas to estimate fetal weight, with no indication of uncertainty. As a result, over half of all high-risk pregnancies go undetected.
THE SOLUTION
The solution applies deep learning to the world’s largest dataset of over 100,000 pregnancies to deliver precise, consistent, and transparent fetal growth assessments. The solution Improves high-risk pregnancy detection by 15%, minimizes operator dependence, and ensures accurate scans even with suboptimal images. This allows midwives with limited training to perform scans with expert-level accuracy. At the same time, uncertainty estimates support clearer and more informed clinical decision
Early detection of high-risk pregnancies using deep learning - Rigshospitalet
Call 4- 2023
500.000 DKK
Clinical Area
Obstetrics
Technology
AI works together with Ultrasound
PROJECT SUMMARY
An AI-based solution that improves detection of high-risk pregnancies by up to15% compared to standard practice. This enhances clinicians’ ability to deliver accurate, transparent, and personalized risk estimates, increasing detection of high-risk pregnancies and allowing non-experts to conduct expert-level care.
CLINICAL IMPACT
The solution improves detection of growth-restricted fetuses, enabling timely interventions and better pregnancy outcomes. It reduces risks of stillbirth and preterm birth, while uncertainty estimates support informed decisions. By enhancing diagnostic accuracy and allowing non-experts to perform expert-level scans, it ensures consistent care, efficient resource use, and significant gains in survival and maternal-child health.