DiffuSETS: 12-lead ECG Generation Conditioned on Clinical Text Reports and Patient-Specific Information

2025-01-10

AI Summary

To address the scarcity and privacy sensitivity of electrocardiogram (ECG) data, as well as weak semantic alignment and unsystematic evaluation in existing generative methods, this work proposes the first diffusion-based framework for generating 12-lead ECGs conditioned jointly on clinical text reports and structured patient metadata. We introduce a novel cross-modal conditional diffusion architecture that integrates BERT-encoded textual descriptions with patient-specific feature embeddings, and incorporate a multi-scale temporal reconstruction loss to enhance signal fidelity and medical plausibility. Furthermore, we establish the first standardized benchmark for ECG generation, evaluating models across three dimensions: signal quality, semantic consistency, and clinical interpretability. Experiments demonstrate significant improvements over state-of-the-art methods across multiple quantitative metrics; expert blind evaluation confirms the clinical validity of generated ECGs; and the framework successfully enables new applications—including data augmentation, medical education case generation, and exploratory analysis of abnormal patterns.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• 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

Abstract

Heart disease remains a significant threat to human health. As a non-invasive diagnostic tool, the electrocardiogram (ECG) is one of the most widely used methods for cardiac screening. However, the scarcity of high-quality ECG data, driven by privacy concerns and limited medical resources, creates a pressing need for effective ECG signal generation. Existing approaches for generating ECG signals typically rely on small training datasets, lack comprehensive evaluation frameworks, and overlook potential applications beyond data augmentation. To address these challenges, we propose DiffuSETS, a novel framework capable of generating ECG signals with high semantic alignment and fidelity. DiffuSETS accepts various modalities of clinical text reports and patient-specific information as inputs, enabling the creation of clinically meaningful ECG signals. Additionally, to address the lack of standardized evaluation in ECG generation, we introduce a comprehensive benchmarking methodology to assess the effectiveness of generative models in this domain. Our model achieve excellent results in tests, proving its superiority in the task of ECG generation. Furthermore, we showcase its potential to mitigate data scarcity while exploring novel applications in cardiology education and medical knowledge discovery, highlighting the broader impact of our work.

Problem

• ECG signal generation
• data privacy
• medical resource constraints

Innovation

• DiffuSETS
• ECG signal generation
• medical data augmentation