ECGFlowCMR: Pretraining with ECG-Generated Cine CMR Helps Cardiac Disease Classification and Phenotype Prediction

2026-01-28

AI Summary

This work proposes ECGFlowCMR, a novel framework addressing the challenges of high annotation costs and data scarcity in cardiac magnetic resonance (CMR) imaging, as well as cross-modal temporal misalignment and limited structural information in electrocardiogram (ECG) signals. By innovatively integrating a phase-aware masked autoencoder (PA-MAE) with a disentangled anatomical-motion deformation flow (AMDF), the method enables high-fidelity synthesis of cine CMR sequences from a single-lead ECG while effectively aligning temporal dynamics across modalities. Experimental results on the UK Biobank and clinical datasets demonstrate that the generated CMR sequences exhibit high anatomical and temporal realism, significantly improving performance in downstream tasks such as cardiac disease classification and phenotypic prediction.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Cardiac Magnetic Resonance (CMR) imaging provides a comprehensive assessment of cardiac structure and function but remains constrained by high acquisition costs and reliance on expert annotations, limiting the availability of large-scale labeled datasets. In contrast, electrocardiograms (ECGs) are inexpensive, widely accessible, and offer a promising modality for conditioning the generative synthesis of cine CMR. To this end, we propose ECGFlowCMR, a novel ECG-to-CMR generative framework that integrates a Phase-Aware Masked Autoencoder (PA-MAE) and an Anatomy-Motion Disentangled Flow (AMDF) to address two fundamental challenges: (1) the cross-modal temporal mismatch between multi-beat ECG recordings and single-cycle CMR sequences, and (2) the anatomical observability gap due to the limited structural information inherent in ECGs. Extensive experiments on the UK Biobank and a proprietary clinical dataset demonstrate that ECGFlowCMR can generate realistic cine CMR sequences from ECG inputs, enabling scalable pretraining and improving performance on downstream cardiac disease classification and phenotype prediction tasks.

Problem

• Cardiac Magnetic Resonance
• Electrocardiogram
• Cross-modal generation
• Data scarcity
• Cardiac disease classification

Innovation

• ECG-to-CMR generation
• Phase-Aware Masked Autoencoder
• Anatomy-Motion Disentangled Flow
• cross-modal pretraining
• cardiac phenotype prediction