Equivariant symmetry-aware head pose estimation for fetal MRI

2025/12/04 by Muthukrishnan, Ramya, Gagoski, Borjan, Lee, Aryn +4
Computer Science · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Face recognition and analysis #Fetal and Pediatric Neurological Disorders #Neonatal and fetal brain pathology

paper · doi:10.48550/arxiv.2512.04890

openalex publication_date 2025/12/04 · openalex created_date 2025/12/06 · openalex updated_date 2026/07/28

Abstract

We present E(3)-Pose, a novel fast pose estimation method that jointly and explicitly models rotation equivariance and object symmetry. Our work is motivated by the challenging problem of accounting for fetal head motion during a diagnostic MRI scan. We aim to enable automatic adaptive prescription of diagnostic 2D MRI slices with 6-DoF head pose estimation, supported by rapid low-resolution 3D MRI volumes acquired before each 2D slice. Existing pose estimation methods struggle to generalize to clinical volumes due to pose ambiguities induced by inherent anatomical symmetries, as well as low resolution, noise, and artifacts. In contrast, E(3)-Pose captures anatomical symmetries and rigid pose equivariance by construction, and yields robust estimates of the fetal head pose. Our experiments on publicly available and representative clinical fetal MRI datasets demonstrate the superior robustness and generalization of our method across domains. Crucially, E(3)-Pose achieves state-of-the-art accuracy on clinical MRI volumes, supporting future clinical translation. Our implementation is publicly available at github.com/MedicalVisionGroup/E3-Pose.

Citations

• Robust Fetal Pose Estimation across Gestational Ages via Cross-Population Augmentation
• 3D Equivariant Pose Regression via Direct Wigner-D Harmonics Prediction
• SpaER: Learning Spatio-temporal Equivariant Representations for Fetal Brain Motion Tracking
• Registration by Regression (RbR): a framework for interpretable and flexible atlas registration
• Learning with 3D rotations, a hitchhiker's guide to SO(3)
• Fully automated planning for anatomical fetal brain MRI on 0.55T
• Equivariant Single View Pose Prediction Via Induced and Restricted Representations
• A Robust and Interpretable Deep Learning Framework for Multi-modal Registration via Keypoints
• Image to Sphere: Learning Equivariant Features for Efficient Pose Prediction
• NeSVoR: Implicit Neural Representation for Slice-to-Volume Reconstruction in MRI
• e3nn: Euclidean Neural Networks
• SVoRT: Iterative Transformer for Slice-to-Volume Registration in Fetal Brain MRI
• Self-Supervised Equivariant Learning for Oriented Keypoint Detection
• Leveraging Equivariant Features for Absolute Pose Regression
• Approximately Equivariant Networks for Imperfectly Symmetric Dynamics
• Leveraging SE(3) Equivariance for Self-Supervised Category-Level Object Pose Estimation
• Vector Neurons: A General Framework for SO(3)-Equivariant Networks
• Equivariant Point Network for 3D Point Cloud Analysis
• Equivariant Filters for Efficient Tracking in 3D Imaging
• Probabilistic 3D Multi-Modal, Multi-Object Tracking for Autonomous Driving
• Handling Object Symmetries in CNN-based Pose Estimation
• An Analysis of SVD for Deep Rotation Estimation
• EPOS: Estimating 6D Pose of Objects with Symmetries
• 6D Object Pose Regression via Supervised Learning on Point Clouds
• General E(2)-Equivariant Steerable CNNs
• Improved Generalization Bounds of Group Invariant / Equivariant Deep Networks via Quotient Feature Spaces
• On Object Symmetries and 6D Pose Estimation from Images
• A Feature Transfer Enabled Multi-Task Deep Learning Model on Medical Imaging
• On the Continuity of Rotation Representations in Neural Networks
• Real-time Deep Pose Estimation with Geodesic Loss for Image-to-Template\n Rigid Registration
• Tensor field networks: Rotation- and translation-equivariant neural networks for 3D point clouds
• Spherical CNNs
• PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes
• Harmonic Networks: Deep Translation and Rotation Equivariance
• V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation
• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• Adam: A Method for Stochastic Optimization
• PROMO: Real‐time prospective motion correction in MRI using image‐based tracking
• Random sample consensus
• FireANTs: Adaptive Riemannian Optimization for Multi-Scale Diffeomorphic Matching

Related

• Fetal Pose Estimation in Volumetric MRI using a 3D Convolution Neural Network
• Fetuses Made Simple: Modeling and Tracking of Fetal Shape and Pose
• Pose-GuideNet: Automatic Scanning Guidance for Fetal Head Ultrasound from Pose Estimation
• FetalDiffusion: Pose-Controllable 3D Fetal MRI Synthesis with Conditional Diffusion Model
• Geometric Transformation Uncertainty for Improving 3D Fetal Brain Pose Prediction from Freehand 2D Ultrasound Videos