Speaker
Description
One of the anesthesiologist’s key tasks during target controlled infusion (TCI) in total intravenous anesthesia (TIVA), where agents such as propofol are administered intravenously, is assessing the depth of hypnosis (DoH). As DoH cannot be measured directly, clinicians rely on indirect indicators such as the Bispectral Index (BIS), derived from EEG signals. However, pharmacokinetic/pharmacodynamic (PK/PD) models implemented in perfusors are population-based and often fail to capture interpatient variability, leading to inaccurate DoH estimation.
To address this limitation, we propose a modelling framework that augments a population-based PK/PD model with a patient-specific residual dynamic component. This component is formulated as a low-order autoregressive model with exogenous input and captures discrepancies between predicted and measured BIS values using real-time clinical data. Model parameters are identified via recursive least-squares with exponential forgetting, enabling online adaptation during surgery while considering the clinically validated population-based model.
Validation with clinical data demonstrates improved BIS prediction accuracy and reduced error, reflecting individual sensitivity to anesthetics. The proposed framework enables personalized, data-driven modelling suitable for closed-loop DoH control, supporting more precise drug delivery, reduced clinician workload, and enhanced patient safety.
