Integrating multidimensional nociceptive-related cortical features for unsupervised assessment of anesthesia states in rats.

Accurate anesthesia monitoring remains challenging because current approaches primarily assess consciousness while overlooking nociceptive processing. Although electroencephalography (EEG)-based metrics such as permutation entropy (PE) and permutation cross-mutual information (PCMI) are widely used, nociceptive-evoked cortical responses, especially gamma-band oscillations (GBOs), a robust index of nociceptive intensity, are rarely incorporated into anesthesia assessment. Here, we recorded electrocorticography from 23 rats under isoflurane anesthesia with nociceptive laser stimulation during both induction and emergence. We extracted GBOs, PE, and PCMI to evaluate their sensitivity to anesthesia states. GBOs and PE tracked nociceptive-related changes during induction, whereas PCMI was more sensitive during emergence. Integrating these multidimensional features, an unsupervised k-means framework identified four latent states: awake, shallow anesthesia, moderate anesthesia, and burst suppression. These findings establish nociceptive-evoked cortical responses as label-free markers and provide a scalable foundation for real-time, closed-loop anesthesia monitoring that jointly assesses consciousness and analgesia.