Clinical Decision Support (CDS) systems in anesthesia are generally limited to simple thresholding and variance detection. Moreover, existing closed-loop control strategies are often not tailored to the practical needs of anesthesiologists. This project aims at exploiting the potentialities of control, optimization and learning methods in order to propose original approaches, allowing to merge the different measured data and to design advanced detection and control systems addressing key clinical challenges. 

A non-exhaustive list of such challenges includes: 

• High uncertainty in biomedical systems: Physiological models are often hypothetical and their parameters are described by probability distributions. 

• Unreliable measurement: Measurement may be affected by drug interactions that compromise interpretability (e.g., the influence of Ketamine on the Bispectral Index), or linked to poorly understood clinical phenomena such as analgesia (absence of pain), which lacks a direct measurable indicator and is often inferred from multiple clinical signals.

• Low persistence of excitation in control inputs:  Drug infusion profiles are subject to various clinical constraints, which limits the identifiability of model parameters.

• Critical events impacting control performance: For instance,  bleeding which can alter drug concentrations in the blood affecting thereby the system dynamics

The Clinical project aims at tackling some of these complexities through time-series classification, probabilistic and stochastic control and Model Predictive Control approaches.  It will exploit both clinical data and available physiological simulators to design more realistic and practice-oriented control approaches that integrate clinicians' feedback.