CardioMorphNet: Cardiac Motion Prediction Using a Shape-Guided Bayesian Recurrent Deep Network

2025-08-28

AI Summary

Accurate estimation of cardiac motion is critical for functional assessment, yet conventional intensity-similarity-based registration methods for cardiac magnetic resonance (CMR) images often neglect anatomical structure, yielding unreliable deformation fields. To address this, we propose a shape-guided Bayesian recurrent deep network that replaces intensity-based losses with recursive registration of cardiac segmentation masks—thereby explicitly focusing on anatomically relevant regions. Our framework integrates a recurrent variational autoencoder, dual-posterior modeling, and 3D deformable registration to explicitly capture spatiotemporal dependencies in short-axis CMR sequences. Furthermore, we introduce a Bayesian loss function that jointly optimizes the motion field and its associated uncertainty map. Evaluated on the UK Biobank dataset, our method achieves significantly improved motion estimation accuracy and anatomical consistency, while producing lower uncertainty and higher confidence specifically within the myocardial region.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Accurate cardiac motion estimation from cine cardiac magnetic resonance (CMR) images is vital for assessing cardiac function and detecting its abnormalities. Existing methods often struggle to capture heart motion accurately because they rely on intensity-based image registration similarity losses that may overlook cardiac anatomical regions. To address this, we propose CardioMorphNet, a recurrent Bayesian deep learning framework for 3D cardiac shape-guided deformable registration using short-axis (SAX) CMR images. It employs a recurrent variational autoencoder to model spatio-temporal dependencies over the cardiac cycle and two posterior models for bi-ventricular segmentation and motion estimation. The derived loss function from the Bayesian formulation guides the framework to focus on anatomical regions by recursively registering segmentation maps without using intensity-based image registration similarity loss, while leveraging sequential SAX volumes and spatio-temporal features. The Bayesian modelling also enables computation of uncertainty maps for the estimated motion fields. Validated on the UK Biobank dataset by comparing warped mask shapes with ground truth masks, CardioMorphNet demonstrates superior performance in cardiac motion estimation, outperforming state-of-the-art methods. Uncertainty assessment shows that it also yields lower uncertainty values for estimated motion fields in the cardiac region compared with other probabilistic-based cardiac registration methods, indicating higher confidence in its predictions.

Problem

• Accurately estimating cardiac motion from CMR images
• Overcoming limitations of intensity-based registration methods
• Providing uncertainty-aware motion field predictions

Innovation

• Bayesian recurrent deep network
• shape-guided deformable registration
• spatio-temporal variational autoencoder