Automatic volumetric measurements of the embryo and head during the first trimester using artificial intelligence: the Rotterdam Periconception Cohort

Non-invasive volumetric measurements during the first trimester of pregnancy provide unique insights into growth and development. Currently, methods for obtaining these measurements, such as semi-automatic (e.g. virtual reality set-up) or tracing methods (e.g. manual segmentation), are time-consuming and not always accurate. To enable routine clinical use of these measurements, we developed an automatic AI algorithm to measure the volume. Here, we focus on the embryonic volume (EV) and head volume (HV) measurements during the first trimester of pregnancy. We utilised three-dimensional ultrasound data sets from the Rotterdam Periconception Cohort, collected between 7 and 11 weeks of gestation. We trained two algorithms on 648 (EV) and 383 (HV) data sets, respectively, and tested them on 164 (EV) and 92 (HV) data sets. We compared the accuracy and time needed between the AI and VR methods. We found an intraclass correlation coefficient of 0.998 for both the EV and HV on the test data set, indicating that the AI algorithm achieved high accuracy, which is supported by the visualisations in figure 1. In terms of time, the AI algorithm took only one minute per data set on average, while experienced raters took 5-10 minutes using virtual reality. We developed an automatic AI algorithm to accurately measure the EV and HV in first trimester ultrasound scans. Since assessing volumetric data now only takes a couple of minutes, the use of these measurements to monitor growth and development, using all three dimensions, during pregnancy becomes feasible. Ultimately, this will contribute to better screening, diagnostics and treatment of developmental disorders in pregnancy. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.