Contactless atrial fibrillation detection system achieves clinical-grade accuracy
A groundbreaking contactless detection system for atrial fibrillation using radio technology and artificial intelligence has demonstrated clinical-grade diagnostic accuracy. Published in Nature Communications, the innovative system records heart mechanical motion via radar sensing and analyses it using neural networks, enabling detection without physical contact and potentially facilitating early diagnosis before symptoms appear.
This technology uses radio signals to remotely capture submillimeter cardiac motion and leverages AI-driven knowledge transfer from ECG diagnostics to identify atrial fibrillation patterns. © Yuqin Yuan and Jinbo Chen
Chinese researchers have developed an innovative contactless system for detecting atrial fibrillation (AF) that could revolutionise early diagnosis of this common and potentially dangerous heart rhythm abnormality.
The system, detailed in a study published in Nature Communications on 20 May 2025, uses millimetre-wave radar technology combined with artificial intelligence to detect the characteristic patterns of AF with sensitivity and specificity comparable to traditional electrocardiogram (ECG) methods.
How the system works
The technology operates by recording heart mechanical motion in a contactless manner through radar sensing. As radio signals reach the body surface, reflections occur which linearly modulate the body surface motion to phase variations in the echo signals. By extracting these phase variations, researchers can recover the corresponding mechanical motion of the heart.
The system then analyses this data using a neural network model that leverages knowledge transfer from ECG diagnostics to enable precise recognition of atrial fibrillation patterns. This approach allows the system to translate cardiac mechanical motions, captured by radio signals, into AF detection in a completely contactless, operation-free, and device-free manner.
Researchers evaluated the system using data from 6,258 outpatients, including 229 with atrial fibrillation, during routine 30-second ECG screenings. The technology demonstrated impressive diagnostic performance with a sensitivity of 0.844 (95% CI, 0.790-0.884) and a specificity of 0.995 (95% CI, 0.993-0.997).
Real-world applications demonstrated
The system’s practical potential was further validated through two key real-world scenarios. In one experiment, the technology was deployed to monitor 27 patients during their sleep routines.
Two subjects were diagnosed with AF by the system prior to their clinical diagnosis, demonstrating its potential for early detection. The researchers found that the reliability of diagnosis improved with monitoring duration, achieving 0.96 reliability (95% CI, 0.93-0.99) with just 42.5 minutes of monitoring data.
In another experiment, researchers monitored five AF patients before and after catheter ablation surgery. The system successfully detected the transition from AF to sinus rhythm states following the procedure, confirming its sensitivity to AF episode transitions.
Addressing a critical healthcare need
Atrial fibrillation is the most common sustained cardiac arrhythmia associated with significant symptoms and health problems. The condition has been rapidly increasing globally, with affected individuals rising from 33.5 million to 59 million between 2010 and 2019.
Traditional ECG-based diagnosis typically involves brief screenings that may miss intermittent episodes, especially during the initial stages of AF. The progression of AF can evolve from initial, sporadic episodes to a persistent form, potentially leading to a permanent and treatment-resistant disease state.
The researchers note that “timely detection and early treatment” are crucial in “mitigating its progression and associated complications, including stroke and heart failure.”
Future developments and limitations
While the current system shows promise, the researchers acknowledge certain limitations. Although the system was only tested in subjects in relatively stationary states, the results suggest it may have potential for practical daily life deployment, aiding early detection and proactive management of atrial fibrillation.
The study authors note: “This study aimed to address the emerging need for early AF diagnosis, especially in the absence of symptoms. Our innovative application of millimetre-wave radar in an AI diagnostic framework offers a contactless, operation-free, device-free experience for AF monitoring. This enables a viable pathway towards lifelong AF monitoring, facilitating the detection of initial AF episodes as early as possible for prompt intervention.”
The technology could potentially be integrated into everyday environments, such as sleep and work settings, providing a feasible pathway toward lifelong proactive monitoring that covers the full spectrum from non-AF to AF progression.
As the authors conclude, this approach “could extend existing AF screening and diagnosis workflows towards a personalised and proactive AF management strategy, ultimately leading to more efficient cardiovascular healthcare.”
Reference
Yuan, Y., Chen, J., Zhang, D., et. al. (2025). Atrial fibrillation detection via contactless radio monitoring and knowledge transfer. Nature Communications, 16(4317). https://doi.org/10.1038/s41467-025-59482-y
