Turning One Wearable ECG Lead into a 12-Lead ECG View

What is it about?

Many wearable devices, such as smartwatches and smart rings, can record only a single ECG lead. However, a standard clinical ECG uses 12 leads, which provide a richer view of the heart’s electrical activity. This paper presents mEcgNet, a compact AI model that reconstructs 12-lead ECG signals from a single Lead-I signal. Instead of using one large neural network, mEcgNet first separates the input ECG into low-, mid-, and high-frequency parts. These parts capture different ECG patterns, such as slower wave shapes and faster heartbeat spikes. The model then processes the frequency parts with small modular neural networks to reconstruct the full 12-lead ECG signal. The goal is to make 12-lead ECG reconstruction more practical for resource-constrained devices, while preserving reconstruction accuracy.

Featured Image

Photo by Joachim Schnürle on Unsplash

Photo by Joachim Schnürle on Unsplash

Why is it important?

Wearable ECG monitoring is becoming increasingly common, but most wearable devices can only measure a limited ECG signal. At the same time, reconstructing richer ECG information with large deep learning models can be too expensive for small, low-power devices. Sending ECG data to the cloud may also raise privacy concerns because ECG signals are sensitive personal health data. mEcgNet addresses this challenge by combining frequency-based ECG signal partitioning with a modular, parameter-efficient neural network design. In experiments on public ECG datasets, mEcgNet achieved up to 23.1× fewer parameters, 5.4× faster inference, and 22.1% lower reconstruction error compared with prior ECG reconstruction models evaluated in the paper. This work could help move richer ECG processing closer to wearable devices, supporting more practical, efficient, and privacy-aware AI systems for digital health.

Perspectives