EMG contributes to improve Cerebral State Index modeling in dogs anesthesia

Cerebral State Index (CSI) is a measure of depth of anesthesia (DoA) developed for humans, which is traditionally modeled with the Hill equation and the propofol effect-site concentration (Ce). The CSI has been studied in dogs and showed several limitations related to the interpretation of EEG data. Nevertheless, the CSI has a lot of potential for DoA monitoring in dogs, it just needs to be adjusted for this species. In this work, an adapted CSI model is presented for dogs considering a) both Ce and EMG as inputs and b) a fuzzy logic structure with parameters optimized using the ANFIS method. The new model is compared with traditional Hill model using data from dogs in routine surgery. The results showed no significant impact in the model performance with the change of model structure (Fuzzy instead of Hill). The residuals of the Hill model were significantly correlated with the EMG, indicating that the latter should be considered in the model. In fact, the EMG introduction in CSI model significantly decreased the modeling error: 11.8 [8.6; 15.2] (fuzzy logic) versus 20.9 [16.4; 29.0] (Hill). This work shows that CSI modeling in dogs can be improved using the current human anesthesia set-up, once the EMG signal is acquired simultaneously with the CSI index. However, it does not invalidate the search of new DoA indices more adjusted to use in dog's anesthesia.