Problem description
The first-trimester structural ultrasound examination (13 weeks of gestation) aims to detect physical abnormalities early in pregnancy. Compared with the 20-week scan, this gives parents and caregivers more time to make difficult decisions when an abnormality is found. A conventional scan needs an experienced sonographer on site and typically takes an half hour. During the scan, the sonographer captures and assesses the planes needed to evaluate the nine fetal structures listed in the national quality standard. The preceding PARADIGM trial showed that a SonoTomography (ST) scan contains enough information for a valid offline assessment of the most important fetal structures. Such a scan consists of a few 3D ultrasound volumes and takes less than five minutes to acquire. The AI system developed in PARADIGM gave reliable measurements of the standard fetal biometry (CRL, HC, AC and FL). However, AI support for assessing the full fetal anatomy is still missing, and so is a clinical workflow for assessing ST scans offline.
Project summary
The ASSET trial builds on these results and investigates whether AI-supported ST scanning can make first-trimester screening more efficient. First, the AI system is extended to all structures in the national protocol. That means that the system will detect all other planes (e.g., the extremities, heart, bladder, etc.) and presents them to the sonographer for an assessment. Second, the project develops a user interface where sonographers can quickly review the selected planes and correct them if needed. The interface automatically records whether each examination is complete, which makes it possible to measure the time saved compared with a conventional assessment. Third, the project designs how results are communicated to the pregnant woman and her partner. Finally, the project starts an image bank of annotated 3D scans of fetal abnormalities, as a basis for future AI that can recognize specific abnormalities. The RTC AI develops the AI models and, together with the Grand Challenge team, the user interface for sonographers.
Funding
This project is funded by The Nijmegen Region Prenatal Screening Foundation (SPN)
