Implicit Neural Representations of Intramyocardial Motion and Strain

2025-09-10

AI Summary

This study addresses the challenge of automatic motion and strain quantification in cardiac MR tagging. We propose an end-to-end displacement field prediction method based on implicit neural representations (INRs), the first to apply INRs to myocardial motion modeling. By integrating conditional latent encoding with a deep regression network, our approach enables optimization-free, continuous, and high-accuracy displacement inference. Trained end-to-end on UK Biobank data, it achieves a 2.14 mm RMSE in tracking accuracy across 452 test cases, attaining the lowest global circumferential and radial strain errors among compared methods; inference speed is 380× faster than the best baseline. Key contributions are: (i) the first INR-based framework specifically designed for cardiac MR tagging; (ii) elimination of iterative optimization, enabling millisecond-scale, continuous strain quantification; and (iii) substantial improvements in clinical applicability and computational efficiency.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Calibration & Uncertainty Quantification
• Intelligent Robots: Motion and Path Planning

Application Category

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

Abstract

Automatic quantification of intramyocardial motion and strain from tagging MRI remains an important but challenging task. We propose a method using implicit neural representations (INRs), conditioned on learned latent codes, to predict continuous left ventricular (LV) displacement -- without requiring inference-time optimisation. Evaluated on 452 UK Biobank test cases, our method achieved the best tracking accuracy (2.14 mm RMSE) and the lowest combined error in global circumferential (2.86%) and radial (6.42%) strain compared to three deep learning baselines. In addition, our method is $sim$380$ imes$ faster than the most accurate baseline. These results highlight the suitability of INR-based models for accurate and scalable analysis of myocardial strain in large CMR datasets.

Problem

• Automatic quantification of intramyocardial motion and strain
• Predict continuous left ventricular displacement without optimization
• Accurate and scalable analysis of myocardial strain in datasets

Innovation

• Implicit neural representations for motion tracking
• Learned latent codes without inference optimization
• Continuous left ventricular displacement prediction