NEF-NET+: Adapting Electrocardio panorama in the wild

TL;DR

Nef-Net+ tackles the challenge of synthesizing panoramic ECG signals from arbitrary viewing angles in real-world conditions by introducing a geometry-aware architecture with Angle Embedding, a View Encoder, and a Geometric View Transformer. It deploys a three-stage development and deployment workflow—Any-Pairs Pretraining, Device Calibration, and On-the-fly Calibration—to address inter-device and patient-specific variability, enabling robust, long-duration panorama synthesis. The Panobench benchmark, a dense 48-view ECG dataset with CT-derived coordinates, provides rigorous evaluation across a wide view spectrum. Experimental results show substantial improvements over prior work, with roughly 6 dB PSNR gains, indicating stronger fidelity and potential clinical utility for panoramic ECG assessment across diverse devices and patient anatomies.

Abstract

Conventional multi-lead electrocardiogram (ECG) systems capture cardiac signals from a fixed set of anatomical viewpoints defined by lead placement. However, certain cardiac conditions (e.g., Brugada syndrome) require additional, non-standard viewpoints to reveal diagnostically critical patterns that may be absent in standard leads. To systematically overcome this limitation, Nef-Net was recently introduced to reconstruct a continuous electrocardiac field, enabling virtual observation of ECG signals from arbitrary views (termed Electrocardio Panorama). Despite its promise, Nef-Net operates under idealized assumptions and faces in-the-wild challenges, such as long-duration ECG modeling, robustness to device-specific signal artifacts, and suboptimal lead placement calibration. This paper presents NEF-NET+, an enhanced framework for realistic panoramic ECG synthesis that supports arbitrary-length signal synthesis from any desired view, generalizes across ECG devices, and compensates for operator-induced deviations in electrode placement. These capabilities are enabled by a newly designed model architecture that performs direct view transformation, incorporating a workflow comprising offline pretraining, device calibration tuning steps as well as an on-the-fly calibration step for patient-specific adaptation. To rigorously evaluate panoramic ECG synthesis, we construct a new Electrocardio Panorama benchmark, called Panobench, comprising 5367 recordings with 48-view per subject, capturing the full spatial variability of cardiac electrical activity. Experimental results show that NEF-NET+ delivers substantial improvements over Nef-Net, yielding an increase of around 6 dB in PSNR in real-world setting. The code and Panobench will be released in a subsequent publication.

Figures (9)

• Figure 2: Our proposed Nef-Net+ architecture for Electrocardio Panorama synthesis (illustrated for a 3-input to 2-query view synthesis task as example). The Nef-Net+ first employs a View Encoder to extract features from the Recorded ECG that are relevant to the Queried ECG. These extracted features are then fused using a Geometric View Transformer to synthesize the query view.
• Figure 3: The Development (Stage I, II, III) and Deployment (Stage IV) Workflow of Nef-Net+.
• Figure 4: Performance of Nef-Net+ Under Varying Input Conditions and Supervision Leads for Reconstruction and Synthesis Tasks.
• Figure 5: Representative examples from CPSC2018 illustrate progressive synthesis of the V5 view by Nef-Net+ across training stages. Green circles mark clinically relevant diagnostic details in the real signals, whereas red circles highlight their increasingly accurate recovery as our multi-stage training progresses.
• Figure 6: Standard ECG waveform with six characteristic components highlighted: P wave, PR segment, QRS complex, ST segment, T wave, and TP segment.