Subscribe to RSS
DOI: 10.1055/s-0041-1739755
A Novel Artificial Intelligence Approach for the Automatic Differentiation of Fetal Occiput Anterior and non-Occiput Anterior Positions during Labor
Objectives To develop a Machine Learning (ML) algorithm for the automatic classification of fetal occiput position at transperineal ultrasound during the second stage of labor.
Methods Multicenter international prospective cohort study including 15 Maternity Hospitals and conducted on singleton term pregnancies with cephalic presenting fetus in the second stage of labor. Firstly, transabdominal ultrasound was performed to assess the fetal occiput position, which was labelled into occiput anterior (OA) or non-OA and represented the gold standard reference for training and validation. Secondly, sonographic images of the fetal head were acquired with TPU on the axial plane and archived on a cloud for remote analysis. A ML-algorithm based on a pattern recognition feed-forward neural network was trained on the transperineal images. In the training phase, the model was trained on labeled data (training dataset), in order to correctly assess the fetal head position, by exploiting geometric, morphological and intensity-based features of the images. In the testing phase, the diagnostic accuracy of the algorithm was evaluated on unlabeled data (testing dataset). Due to the unbalanced numbers of OA and non-OA classes, we also evaluated the algorithm’s performance using the F1-score and Precision-Recall Area Under the Curve (PR-AUC). The Cohen’s kappa (k) evaluated the agreement between the ML-algorithm and the gold standard.
Results Over a period of 24 months, 1219 women in the second stage of labor were enrolled. They were classified as OA (n=801 or 65.7%) or non-OA (n=418 or 34.3%) on the basis of transabdominal ultrasound. From both the sub-groups (OA and non-OA), 70% of the patients were randomly assigned to the training dataset (824 patients) while the remaining 30% (395 patients) were used as testing dataset.On the latter group the ML based algorithm yielded a correct classification of the fetal occiput position in 90.6% of cases (357 out of 395), including 224 out of 246 OA (91.0%) and of 133 out of 149 non-OA images (89.3%). Moreover, the F1-score was 88.7% and PR-AUC was 85.4%. The algorithm showed a balanced performance in the recognition of both anterior and non-anterior occiput positions. A high agreement between the ML-algorithm and the gold standard method was also noted (k=0.81; p<0,0001).
Conclusion A ML-based algorithm for the automatic assessment of the fetal head position at TPU has been developed and can accurately differentiate between OA and non-OA positions. This algorithm has the potential to support not only obstetricians, but also midwives and accoucheurs in the clinical use of TPU. Future studies will specifically address the repeatability and reproducibility of the measurements, as well as the possibility of employing a similar approach to effectively distinguish the different types of non-OA positions.
Publication History
Article published online:
26 November 2021
© 2021. Thieme. All rights reserved.
Georg Thieme Verlag
Rüdigerstraße 14, 70469 Stuttgart,
Germany