Complementarity-Preserving Generative Theory for Multimodal ECG Synthesis: A Quantum-Inspired Approach

Timothy Oladunni; Farouk Ganiyu-Adewumi; Clyde Baidoo; Kyndal Maclin

eess.SPMar 16, 2026

Complementarity-Preserving Generative Theory for Multimodal ECG Synthesis: A Quantum-Inspired Approach

Authors: Timothy Oladunni, Farouk Ganiyu-Adewumi, Clyde Baidoo, Kyndal Maclin

Organizations: Department of Computer Science, Morgan State University, Baltimore, MD, USA

Abstract

Multimodal deep learning has substantially improved electrocardiogram (ECG) classification by jointly leveraging time, frequency, and time-frequency representations. However, existing generative models typically synthesize these modalities independently, resulting in synthetic ECG data that are visually plausible yet physiologically inconsistent across domains. This work establishes a Complementarity-Preserving Generative Theory (CPGT), which posits that physiologically valid multimodal signal generation requires explicit preservation of cross-domain complementarity rather than loosely coupled modality synthesis. We instantiate CPGT through Q-CFD-GAN, a quantum-inspired generative framework that models multimodal ECG structure within a complex-valued latent space and enforces complementarity-aware constraints regulating mutual information, redundancy, and morphological coherence. Experimental evaluation demonstrates that Q-CFD-GAN reduces latent embedding variance by 82%, decreases classifier-based plausibility error by 26.6%, and restores tri-domain complementarity from 0.56 to 0.91, while achieving the lowest observed morphology deviation (3.8%). These findings show that preserving multimodal information geometry, rather than optimizing modality-specific fidelity alone, is essential for generating synthetic ECG signals that remain physiologically meaningful and suitable for downstream clinical machine-learning applications.

Figures & tables

RQ	Objective	Evidence	Key Finding
RQ1	Does Q–CFD–GAN preserve the statistical structure of real ECG signals better than baseline generative models?	Table II ( $\sigma^{2}$ , $\Delta$ , $\rho_{\Delta}$ )	Q–CFD–GAN achieves the lowest variance deviation ( $\sigma^{2}=0.033$ ) and smallest structural error ( $\Delta=0.157$ ), indicating improved statistical fidelity.
RQ2	Does Q–CFD–GAN better preserve multimodal complementarity across ECG domains?	Table II ( $\widehat{C}_{TFS}/C_{TFS}$ )	Q–CFD–GAN restores 91% of real multimodal complementarity, outperforming MI–GAN (0.56) and Q–Base (0.68).
RQ3	Does interference-aware latent modeling improve multimodal synthesis quality?	Table II ( $\gamma$ , $\kappa$ )	Q–CFD–GAN achieves the highest constructive interference score ( $\gamma=1.625$ ) and lowest redundancy coefficient ( $\kappa=0.339$ ).
RQ4	Do generated ECG signals preserve physiologically meaningful amplitude characteristics?	Fig. 3	The peak-to-peak amplitude distribution of CFD-Syn closely matches real ECG signals, while NonCFD-Syn collapses to unrealistic ranges.
RQ5	Do generated signals maintain realistic spectral complexity?	Fig. 5	CFD-Syn preserves spectral entropy characteristics of real ECG signals, while NonCFD-Syn shows entropy collapse.
RQ6	Do generated signals maintain realistic signal energy characteristics?	Fig. 4	CFD-Syn follows the RMS energy distribution of real ECG signals, indicating physiologically plausible signal energy levels.

TABLE I: Research Questions Addressed by Quantitative Evaluation

Framework	Key Limitation for Multimodal ECG Generation
Classical GANs [ 1 , 2 , 3 ]	Single-domain generation with no mechanism for synchronized tri-domain (T/F/S) coherence.
InfoGAN and MI-based GANs [ 16 ]	Encourage disentanglement but do not regulate redundancy, orthogonality, or ECG-specific complementarity.
Multimodal ECG classifiers [ 6 , 7 , 17 , 8 ]	Discriminative models only; incapable of synthesizing coherent multimodal ECG signals.
Complementarity-based analysis frameworks	Quantify multimodal relationships but do not enforce complementarity during data generation.
Quantum-inspired models [ 15 , 13 , 14 ]	Capture phase and interference effects but lack physiology-aware constraints and ECG-specific multimodal structure.
Cardiac simulation models [ 18 , 19 ]	Biophysically detailed but rigid; not aligned with learned information geometry or data-driven multimodal coupling.

TABLE II: Summary Comparison of Q–CFD–GAN with Prior Frameworks

\mathcal{L}_{G}=\mathcal{L}_{\mathrm{GAN}}+\lambda_{1}\mathcal{L}_{\mathrm{CFD}}+\lambda_{2}\mathcal{L}_{\mathrm{interf}}+\lambda_{3}\mathcal{L}_{\mathrm{phys}}.

Algorithm 1 Q–CFD–GAN Training Procedure

Panel of Figure 4 — Fig. 1: Q–CFD–GAN generator architecture. A shared complex-valued latent state is mapped to a common latent core and decoded in parallel to generate synchronized time-, frequency-, and time–frequency ECG representations.

Fig. 2: Workflow for ECG synthesis evaluation using Q–CFD–GAN. Synthetic time, frequency, and time–frequency ECG modalities and real test data are processed through a pretrained multimodal encoder; CFD and morphology metrics jointly quantify multimodal coherence and physiological fidelity.

r_{\sigma}(m),\quad\rho_{\Delta}(m),\quad\gamma(m),\quad\kappa(m).

Algorithm 2 Evaluation of Q–CFD–GAN

Model	$\sigma^{2}$	$r_{\sigma}$	$\Delta$	$\rho_{\Delta}$	$\widehat{C}_{TFS}/C_{TFS}$	$\gamma$	$E_{\mathrm{RMS}}$	$\kappa$
Real ECG	0.031	–	–	–	1.00	–	–	–
MI-GAN	0.042	1.000	0.214	0	0.56	1.000	11.2%	1.000
Q-Base	0.039	0.727	0.187	0.126	0.68	1.214	9.6%	0.857
Q–CFD–GAN	0.033	0.182	0.157	0.266	0.91	1.625	3.8%	0.339

TABLE III: Quantitative Evaluation of Q–CFD–GAN

Fig. 3: Peak-to-peak (P2P) amplitude distribution.

Multimodal Generation

Multimodal generation creates coordinated text, images, audio, video, 3D, and other outputs from mixed inputs while preserving meaning and consistency.

See all 243 papers

Feb 24, 2026

CG-DMER: Hybrid Contrastive-Generative Framework for Disentangled Multimodal ECG Representation Learning

Accurate interpretation of electrocardiogram (ECG) signals is crucial for diagnosing cardiovascular diseases. Recent multimodal approaches that integrate ECGs with accompanying clinical reports show strong potential, but they still face two main concerns from a modality perspective: (1) intra-modality: existing models process ECGs in a lead-agnostic manner, overlooking spatial-temporal dependencies across leads, which restricts their effectiveness in modeling fine-grained diagnostic patterns; (2) inter-modality: existing methods directly align ECG signals with clinical reports, introducing modality-specific biases due to the free-text nature of the reports. In light of these two issues, we propose CG-DMER, a contrastive-generative framework for disentangled multimodal ECG representation learning, powered by two key designs: (1) Spatial-temporal masked modeling is designed to better capture fine-grained temporal dynamics and inter-lead spatial dependencies by applying masking across both spatial and temporal dimensions and reconstructing the missing information. (2) A representation disentanglement and alignment strategy is designed to mitigate unnecessary noise and modality-specific biases by introducing modality-specific and modality-shared encoders, ensuring a clearer separation between modality-invariant and modality-specific representations. Experiments on three public datasets demonstrate that CG-DMER achieves state-of-the-art performance across diverse downstream tasks.

Ziwei Niu, Hao Sun, Shujun Bian +4

Department of Biomedical Engineering, National University of Singapore, Singapore, Singapore · College of Computer Science and Technology, Zhejiang University, Hangzhou, China · College of information science and engineering, Ritsumeikan university, ibaraki Osaka, Japan +1

Electrocardiogram Representation Learning Multimodal Disentangled Representation Learning Representation Learning Electrocardiogram Interpretation Electrocardiography Multimodal Contrastive Learning

Ask PDF

May 26, 2026

Information-theoretic Multimodal Representation Learning for Electrocardiogram Signals

Electrocardiograms (ECGs) are widely used non-invasive measurements of cardiac activity and play a central role in clinical diagnosis. Recent multimodal approaches align ECG signals with clinical reports to incorporate diagnostic semantics, but clinical reports often fail to preserve the rich physiological structure of ECG waveforms, particularly across multiple levels of abstraction ranging from coarse diagnostic categories to fine-grained morphology. To address this limitation, we formulate ECG representation learning from an information-theoretic perspective and derive a tractable objective that jointly preserves signal structure and integrates clinical semantics. Based on this principle, we propose MERIT (Multimodal ECG Representation via Information Theory), a dual-branch pretraining framework combining masked ECG modeling with ECG--text contrastive alignment. Extensive experiments on PTB-XL and additional benchmarks demonstrate consistent improvements over prior methods, including gains exceeding

3%

F1 on PTB-XL All and

5%

F1 on SubClass classification. In zero-shot evaluation, MERIT further improves performance by up to

+2.66\%

AUC and

+2.11\%

F1 on PTB-XL SubClass, while also demonstrating robustness under multiple distribution-shift settings. Moreover, leveraging the learned ECG representations for ECG-conditioned clinical text generation with large language models improves text quality across several metrics, including ROUGE and METEOR. Together, these results demonstrate that MERIT learns more informative and clinically meaningful ECG representations, particularly for fine-grained clinical applications.

Phu X. Nguyen, Konstantinos Kontras, Wei Dai +5

1KU Leuven · 3MIT · 4DFKI +1

Representation Learning Cross-Modal Representation Learning Self-Supervised Time Series Representation Learning Multimodal Pretraining Electrocardiogram Representation Learning Electrocardiogram Classification+2

Ask PDF

Jan 28, 2026

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

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.

Xiaocheng Fang, Zhengyao Ding, Guangkun Nie +9

State Key Laboratory of General Artificial Intelligence, Peking University · School of Intelligence Science and Technology, Peking University, Beijing, China · Polytechnic Institute of Zhejiang University, Zhejiang University, Hangzhou, China +4

Synthetic Data Pretraining Medical Diagnosis Clinical Prediction Medical Image Synthesis Conditional Generative Modeling Cardiac MRI

Ask PDF

Abstract

Multimodal Generation

Explore similar work

CG-DMER: Hybrid Contrastive-Generative Framework for Disentangled Multimodal ECG Representation Learning

Information-theoretic Multimodal Representation Learning for Electrocardiogram Signals

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