Fetal MRI-based artificial intelligence in gestational age prediction——a practical solution to an unsolved problem?

Excerpt

We would like to use the opportunity to comment on the article by Kojita et al, which we have read with great interest [1]. In this study, the authors have chosen a fetal head-MRI–based deep learning approach to predict gestational age in the second, and especially third, trimester of pregnancy. By using first trimester ultrasound—which is per se accurate within 3 to 5 days—as a standard of reference, their model reached an (added) upper limit of 1.66 weeks of accuracy, which is in the accuracy range of sonography-based age prediction in the second trimester (7–14 days) and clearly better than standard ultrasound-based predictions after 28 gestational weeks (> 21 gestational days) [2]. While the presented approach evidently offers benefits in dating pregnancies in the third trimester, some practical considerations need to be taken into account when evaluating its clinical applicability. …