A deep learning algorithm for automatic 3D segmentation and classification of the sheep placenta in magnetic resonance images

Magnetic resonance imaging (MRI) is a powerful non-invasive method for assessing placental morphology and physiology in vivo, offering higher resolution and functional insight than ultrasound. However, placental MRI analysis is limited by manual segmentation and time-consuming classification workflows that are prone to inter-operator variability in preclinical sheep models. This study presents the first fully automated MRI analysis pipeline for the ovine placenta, integrating advanced deep learning models to accelerate and standardize preclinical placental assessment. We evaluated 2D and 3D configurations of the self-configuring nnU-Net against a conventional 2D U-Net, training on expert-annotated datasets and testing on a held-out set. The 3D nnU-Net achieved a Dice score above 0.81, outperforming both 2D approaches while reducing processing time by more than 95% and closely matching manual contours. Using these segmentations, we implemented a YOLOv11-based detection module to localize and classify individual placentomes into four morphological types (A-D), achieving 88.7% precision, 92.1% recall, mAP50 of 95.1%, and mAP50-95 of 93.5%. This dual-stage workflow enables fully automated, in vivo quantification of placentome morphology in sheep, supporting high-throughput, reproducible assessment of placental structural and functional adaptation while reducing operator bias and manual burden, and laying the foundation for MRI biomarker validation.