To identify congenital anomalies, pregnant women are offered a structural screening using ultrasound during the first trimester. This screening is performed in two-dimensional ultrasound and has a duration of up to 45 minutes. While three-dimensional (3D) and 3D virtual reality (VR) ultrasound have shown to be beneficial for the anomaly detection rate of some fetal structures, their implementation is limited by an increased evaluation time and the need for expert knowledge. To address these problems and reduce operator dependency, an automatic organ detection model for first-trimester fetal ultrasound is proposed. We used 69 3D ultrasound scans from the VR-FETUS study with a gestational age ranging from 11+0 weeks and 13+6 weeks. These scans were annotated in VR with sparse labels (landmarks) for the heart, lungs, spine, choroid plexuses, cerebellum, mandible, orbits, upper lip and nasal bone, while dense labels (segmentations) were provided for the bladder, kidneys and stomach. In total 50 scans were used for training and 19 scans were used for testing. We trained nnU-Net models for each organ separately, using pseudo segmentation labels that were created from the landmarks. Furthermore, it was tested if using a combination of labels, additional training data or different loss functions would increase model performance. The Dice similarity coefficient (DSC) was used for evaluation of the segmentation labels and we assessed the detection rate for all models. The heart, lungs and mandible were detected in 95 to 100% of the test scans, while the cerebellum and plexuses were detected in 63 to 79% of the test scans. The detection rate for the orbits and upper lip was lower than 43% and the nasal bone was not detected by any model. The median DSC for the bladder, kidneys and stomach ranged between 0.70 and 0.81 and they were detected in 89 to 100% of the test scans. With our results we take the first step towards automatic organ detection in first-trimester 3D ultrasound, by lowering the evaluation time and reducing the operator dependency.

The periconceptional period, encompassing the embryonic phase, is a critical window where a majority of reproductive failures, pregnancy complications, and adverse pregnancy outcomes arise. The Carnegie staging system comprises 23 stages which are based on embryonic morphological development. This allows for the assessment of normal and abnormal embryonic development during this critical period. In-utero Carnegie staging using three-dimensional (3D) ultrasound scans visualized with virtual reality offers valuable insights but is currently a time-consuming manual process. To address this, we propose a deep learning approach for Carnegie staging in 3D ultrasound scans.

We used a dataset comprising 1413 3D ultrasound scans from the Rotterdam Periconceptional Cohort, annotated with Carnegie stages spanning from stages 13 to 23, including fetal subjects. Various training strategies were explored. We compared a metric regression approach, which considers the ordered nature of the Carnegie stages by treating the Carnegie stages as a continuous variable, with a multi-class classification approach, treating stages as independent categories. Additionally, we evaluated the influence of using a loss function accommodating the categorical nature of the Carnegie stages in the metric regression approach and examined the impact of incorporating embryonic size in the model input. Ultimately, a regression approach using the Mean Squared Error (MSE) loss function emerged as the optimal choice.

This model achieved a classification accuracy of 0.59 and a Root Mean Squared Error (RMSE) of 0.62 on the test set. This performance is comparable to an intermediate human rater, which achieved an accuracy of 0.63 and a RMSE of 0.65. Our findings represent a significant step towards the development of an automated Carnegie staging method, offering the potential for a more comprehensive evaluation of the critical embryonic phase in the clinic.