Scholarlink

<p>Electrocardiograms (ECGs) are essential, non-invasive diagnostic tools for assessing cardiac conditions. Existing methods often have limited generalizability, focus on narrow condition sets, and rely on raw physiological signals, which may be unavailable in resource-limited settings where only printed or digital ECG images are accessible. Recent advances in multimodal large language models (MLLMs) offer new opportunities, yet ECG image interpretation remains challenging due to the lack of instruction-tuning data and standardized benchmarks. To address these gaps, we introduce <Emphasis FontCategory="NonProportional">ECGInstruct</Emphasis>, the first large-scale ECG image instruction-tuning dataset with over one million samples, covering diverse tasks including feature recognition, rhythm analysis, morphology assessment, and clinical report generation. We develop <Emphasis FontCategory="NonProportional">PULSE</Emphasis>, a fully open-source MLLM for ECG image interpretation trained on <Emphasis FontCategory="NonProportional">ECGInstruct</Emphasis>. We further curate <Emphasis FontCategory="NonProportional">ECGBench</Emphasis>, a human expert-developed benchmark spanning four core ECG interpretation tasks across nine datasets, incorporating both synthesized and real-world ECG images to enable clinically realistic evaluation. Our experiments demonstrate that <Emphasis FontCategory="NonProportional">PULSE</Emphasis>establishes a new state of the art, outperforming general-purpose MLLMs by 21% to 33% in average accuracy. These results highlight the potential of <Emphasis FontCategory="NonProportional">PULSE</Emphasis>to improve ECG image interpretation in clinical practice. All code, data and models are available at <a href="https://aimedlab.github.io/PULSE/">https://aimedlab.github.io/PULSE/</a>.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Teaching multimodal LLMs to comprehend 12-lead electrocardiographic images

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月16日

查看全文
  • Ruoqi Liu,
  • Yuelin Bai,
  • Xiang Yue,
  • Ping Zhang

摘要

Electrocardiograms (ECGs) are essential, non-invasive diagnostic tools for assessing cardiac conditions. Existing methods often have limited generalizability, focus on narrow condition sets, and rely on raw physiological signals, which may be unavailable in resource-limited settings where only printed or digital ECG images are accessible. Recent advances in multimodal large language models (MLLMs) offer new opportunities, yet ECG image interpretation remains challenging due to the lack of instruction-tuning data and standardized benchmarks. To address these gaps, we introduce ECGInstruct, the first large-scale ECG image instruction-tuning dataset with over one million samples, covering diverse tasks including feature recognition, rhythm analysis, morphology assessment, and clinical report generation. We develop PULSE, a fully open-source MLLM for ECG image interpretation trained on ECGInstruct. We further curate ECGBench, a human expert-developed benchmark spanning four core ECG interpretation tasks across nine datasets, incorporating both synthesized and real-world ECG images to enable clinically realistic evaluation. Our experiments demonstrate that PULSEestablishes a new state of the art, outperforming general-purpose MLLMs by 21% to 33% in average accuracy. These results highlight the potential of PULSEto improve ECG image interpretation in clinical practice. All code, data and models are available at https://aimedlab.github.io/PULSE/.