Consciousness-ECG Transformer for Conscious State Estimation System with Real-Time Monitoring

TL;DR

This work tackles the challenge of non-invasively estimating conscious states in clinical settings where EEG is impractical due to noise and environmental constraints. It introduces the Consciousness-ECG Transformer, a two-branch architecture with a Temporal Encoder and a Cardiac Cycle Encoder that employ decoupled query attention and rotary positional encoding to capture both cardiac-cycle morphology and inter-beat HRV. The approach achieves state-of-the-art performance on overnight sleep staging and anesthesia-level monitoring in animals, with ACC up to 0.880, MF1 up to 0.845, and AUC up to 0.895, while enabling real-time mobile deployment and low latency. The work demonstrates the practical viability of ECG-based conscious-state monitoring in dynamic clinical environments and discusses clinical applicability, ablation findings, and directions for future improvement.

Abstract

Conscious state estimation is important in various medical settings, including sleep staging and anesthesia management, to ensure patient safety and optimize health outcomes. Traditional methods predominantly utilize electroencephalography (EEG), which faces challenges such as high sensitivity to noise and the requirement for controlled environments. In this study, we propose the consciousness-ECG transformer that leverages electrocardiography (ECG) signals for non-invasive and reliable conscious state estimation. Our approach employs a transformer with decoupled query attention to effectively capture heart rate variability features that distinguish between conscious and unconscious states. We implemented the conscious state estimation system with real-time monitoring and validated our system on datasets involving sleep staging and anesthesia level monitoring during surgeries. Experimental results demonstrate that our model outperforms baseline models, achieving accuracies of 0.877 on sleep staging and 0.880 on anesthesia level monitoring. Moreover, our model achieves the highest area under curve values of 0.786 and 0.895 on sleep staging and anesthesia level monitoring, respectively. The proposed system offers a practical and robust alternative to EEG-based methods, particularly suited for dynamic clinical environments. Our results highlight the potential of ECG-based consciousness monitoring to enhance patient safety and advance our understanding of conscious states.

Figures (8)

• Figure 1: The overall conscious state estimation system with real-time monitoring, which includes the mobile ECG monitoring device, temporal encoder, cardiac cycle encoder, and classification head.
• Figure 2: Architecture of temporal encoder and ConvBlock.
• Figure 3: Architecture of cardiac cycle encoder and decoupled query attention; RoPE represents the rotary positional embedding.
• Figure 4: (a) Setting of developed conscious state estimation system in overnight sleep staging. SIG, REF, and GND represent the signal, reference, and ground electrodes, respectively. (b) Setting of developed conscious state estimation system during anesthesia level monitoring. (c) Example of hypnogram from overnight sleep staging. N1, N2, and N3 represent non-rapid eye movement sleep stage 1, 2, 3, respectively and REM indicates rapid-eye movement sleep. (d) Example of medication administration record of sedative and antagonist from anesthesia level monitoring during surgery.
• Figure 5: Comparison of ROC curves and AUC values across various methods.