Autonomic Signature-Driven Anesthesia Depth Monitoring with Biomimetic Wearable ECG and Knowledge Graph-Augmented Deep Networks.

Considerable efforts have been devoted to accurately monitoring the depth of anesthesia to ensure patient safety during surgery. Traditional approaches typically rely on electroencephalogram (EEG)-based indices, such as the Bispectral Index (BIS), which require specialized equipment. In contrast, electrocardiogram (ECG) signals are widely available in clinical settings and can be conveniently acquired via wearable devices, while also exhibiting strong responsiveness to anesthetic agents. Inspired by biomimetic physiological regulation mechanisms, this study proposes a wearable-compatible ECG-based framework for depth-of-anesthesia detection that leverages autonomic nervous system characteristics and a knowledge graph-enhanced graph convolutional network (GCN). ECG recordings from 110 patients were preprocessed, and 20 anesthesia-related features were extracted, spanning morphological, statistical, spectral, heart rate variability (HRV), and entropy-based descriptors; feature selection methods identified 13 discriminative features. A patient-level knowledge graph was first constructed using the 88 training patients (1760 nodes), and test patient nodes were incorporated only after training was complete for inductive inference. Experimental results demonstrate that the proposed deep knowledge GCN achieves a test accuracy of 98.18% in distinguishing between awake and deep sleep anesthesia states, indicating that biomimetic, wearable-compatible ECG analysis combined with knowledge graph learning holds strong potential as a cost-effective alternative to traditional EEG-based anesthesia monitoring systems.