ECGomics: An Open Platform for AI-ECG Digital Biomarker Discovery

TL;DR

ECGomics reframes ECG analysis as a multidimensional, omics-inspired platform to extract digital biomarkers. It combines expert-driven structural/intensity/functional features with deep representations via ECGFounder to bridge interpretability and accuracy. The platform includes web and mobile tools enabling high-throughput biomarker extraction and real-time reporting, validated across AF detection, AF recurrence, occult coronary disease screening, and maternal health contexts. This approach advances precision cardiovascular medicine by enabling systemic health insights from ECG data while highlighting remaining challenges in data scale, annotation quality, and multi-center validation.

Abstract

Background: Conventional electrocardiogram (ECG) analysis faces a persistent dichotomy: expert-driven features ensure interpretability but lack sensitivity to latent patterns, while deep learning offers high accuracy but functions as a black box with high data dependency. We introduce ECGomics, a systematic paradigm and open-source platform for the multidimensional deconstruction of cardiac signals into digital biomarker. Methods: Inspired by the taxonomic rigor of genomics, ECGomics deconstructs cardiac activity across four dimensions: Structural, Intensity, Functional, and Comparative. This taxonomy synergizes expert-defined morphological rules with data-driven latent representations, effectively bridging the gap between handcrafted features and deep learning embeddings. Results: We operationalized this framework into a scalable ecosystem consisting of a web-based research platform and a mobile-integrated solution (https://github.com/PKUDigitalHealth/ECGomics). The web platform facilitates high-throughput analysis via precision parameter configuration, high-fidelity data ingestion, and 12-lead visualization, allowing for the systematic extraction of biomarkers across the four ECGomics dimensions. Complementarily, the mobile interface, integrated with portable sensors and a cloud-based engine, enables real-time signal acquisition and near-instantaneous delivery of structured diagnostic reports. This dual-interface architecture successfully transitions ECGomics from theoretical discovery to decentralized, real-world health management, ensuring professional-grade monitoring in diverse clinical and home-based settings. Conclusion: ECGomics harmonizes diagnostic precision, interpretability, and data efficiency. By providing a deployable software ecosystem, this paradigm establishes a robust foundation for digital biomarker discovery and personalized cardiovascular medicine.

Figures (6)

• Figure 1: The taxonomic parallel of ECGomics compared with Genomics. The framework establishes a systematic analogy between Genomic and ECGomics to redefine ECG as a high-throughput omics resource. (Top) Genomics deconstructs DNA information into Structural Genomics, Mutational Genomics, Functional Genomics, and Comparative Genomics. (Bottom) Correspondingly, ECGomics deconstructs cardiac signals into four dimensions: Structural ECGomics, Intensity ECGomics, Functional ECGomics, and Comparative ECGomics. This multi-dimensional mapping enables the translation of raw electrical signals into a digital biomarker for systemic health assessment and disease trajectory prediction.
• Figure 2: The ECGomics-driven workflow: from digital biomarker to clinical translation. This schematic outlines the end-to-end architecture of the ECGomics paradigm, positioning it as a pivotal nexus that connects cardiovascular phenomics with systemic biology. The integrative framework is organized into three hierarchical layers. ECGomics Data Layer (Left): This layer facilitates high-throughput, multi-modal feature extraction from raw ECG signals. By employing structural, intensity, functional, and comparative ECGomics analyses, it deconstructs complex waveforms into structured digital biomarkers. Biomedical Analysis Layer (Middle): Serving as an integrative hub, this layer correlates ECG-derived signatures with demographics, laboratory biomarkers, multi-modal imaging, and vital signs. It enables deep phenotyping and multi-omics association studies (including genomics, transcriptomics, proteomics, and metabolomics) to support predictive tasks, such as diagnostic classification, prognostic stratification, and clinical decision support. Application Layer (Right): This stage translates high-dimensional insights into actionable clinical pathways, encompassing precision behavioral interventions (e.g., sleep and exercise management), personalized therapeutics, and the dynamic modeling of disease trajectories. Ultimately, this framework exemplifies an integrative paradigm that bridges the gap between cardiovascular phenotypic data and multi-omics information, facilitating advanced precision health management.
• Figure 3: AI-ECG Digital Biomarker overview. Illustrates the classification logic from the underlying raw features to the high-level predictive model, mainly including: engineered biomarkers (corresponding to structural, intensity and functional dimensions, such as waveform duration, spectrum and heart rate variability), predictive biomarkers (corresponding to comparison dimensions, such as cardiac age and disease risk prediction), and deep biomarkers based on deep learning (such as various hidden layer embedded features).
• Figure 4: The user interface of the ECGomics platform and usage steps are displayed on the right side. (A) Allows for the configuration of signal acquisition settings, including SampleRate (Hz), adcGain, and adcZero (mV), ensuring the raw data is correctly calibrated for subsequent processing. (B) Select Sample Data from a dropdown menu or upload custom .npy files for analysis. (C) Shows the raw numerical representation of the ECG signal, allowing for immediate verification of the digital data string before visualization. (D) Provides a high-fidelity visualization of the 12-lead ECG waveform. It includes a standard grid for clinical assessment and an Export function for extracting digitized waveform data. (E) Positioned at the bottom of the left panel, this Generate ECGomics control serves as the execution command to initiate the AI-driven biomarker extraction process. (F) A utility feature that allows users to switch the interface language between Chinese and English, facilitating international research collaboration. (G) Represents the core output of the platform, categorized into the four distinct dimensions of the ECGomics taxonomy. All the analysis results can be exported in CSV format by clicking the green icon on the right.
• Figure 5: Operation Process for Portable ECG Collection and ECGomics Analysis. (A) Navigate to the primary mobile application dashboard and execute the "Start Collection" command to initialize the automated data acquisition and processing sequence. (B) Securely hold the handheld portable ECG transducer with both hands to establish a stable connection; the interface provides real-time, synchronous visual feedback of the collection progress (e.g., the countdown and live waveform), ensuring the integrity of the captured electrocardiographic signals. (C) Upon successful completion of the recording, the platform performs near-instantaneous processing to deliver a comprehensive analysis report. This output facilitates a granular assessment of cardiac health by integrating diverse interpretive dimensions, including high-fidelity waveform visualization, automated rhythm diagnostics (e.g., Heart Rate, PR interval, QRS duration), and advanced predictive metrics such as biological heart age and cardiovascular risk stratification.