Key result
Radar-based deep learning fusion model detects heartbeats with ~97% F1 accuracy, outperforming single-signal approaches.
Why the study?
Continuous wave radar offers a contactless alternative to ECG for heart rate and heart rate variability monitoring, but optimal signal fusion methods for accurate assessment are unclear.
Does a 61 GHz continuous wave radar system using an attention-based deep learning fusion model accurately monitor heart rate and heart rate variability contactlessly?
Observational (n=20)
Does a 61 GHz continuous wave radar system using an attention-based deep learning fusion model accurately monitor heart rate and heart rate variability contactlessly?
A 61 GHz continuous wave radar system combined with an attention-based deep learning fusion model demonstrates high accuracy for contactless heart rate and heart rate variability monitoring.
Supports feasibility of contactless HR/HRV monitoring; leaves open validation in larger clinical cohorts.
Continuous wave (CW) radar is an alternative to gold-standard electrocardiography (ECG) for contactless heart rate (HR) monitoring and heart rate variability (HRV) analysis. This work aims to monitor HR and HRV using a deep learning (DL) model with a 61 GHz CW radar system. We employed attention-based DL fusion model to combine information from radar heart sound (HS) and radar pulse wave (PW) signals. Additionally, we compared attention-based fusion with alternative fusion methods, including data fusion, feature fusion, and decision fusion. To evaluate the performance of the proposed method, a dataset of approximately 20 hours was collected from 20 participants in varying body positions. Using attention-based fusion, we achieved the overall highest F1 score of 97.15% for heartbeat detection, which outperformed solely HS-based or PW-based model, as well as other fusion methods. F1 scores exceeding 96% were obtained for all analyzed body positions. A high correlation of 97.11% and an RMSE of 28.99 ms were achieved for the estimation of the interbeat interval (IBI). In addition, we extracted HRV in time, frequency, and nonlinear domains. Low relative errors were achieved, e.g. 14.04% for TRI, 6.14% for LF, 6.35% for alpha2, and 10.93% for ApEn. The proposed approach achieved high-level performance for HR monitoring and HRV estimation using a CW radar system, showing great potential for various healthcare applications.
No takes yet. Share an insight, caveat, or question.
Lu et al. (2026) reported an observational. Attention-based deep learning fusion model with 61 GHz continuous wave radar vs. Single-signal models and alternative fusion methods was evaluated on Heartbeat detection F1 score. An attention-based deep learning fusion model using a 61 GHz continuous wave radar achieved a 97.15% F1 score for heartbeat detection, outperforming alternative models.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: