Back to main page of IDEA

IDEA Project Summary

In the last 10 years major efforts have been made in the development of new tools to monitor the brain activity for evaluating the depth of anaesthesia. The introduction of the Bispectral index from the electroencephalogram (BIS) or the entropy measure (EM) for measuring the degree of unconsciousness are examples of such improvements. These parameters complement the use of autonomic responses which used alone do not adequately reflect the level of consciousness and may lead to decisions that increase the risk of anaesthesia. The existence of adequate monitors (sensors) like BIS and EM and actuators allows the automation of anaesthesia, motivated by the following issues:
• The anaesthetist is relieved from repetitive tasks and has more capacity to concentrate on patient supervision and on the decisions requiring expert reasoning capabilities;
• Automation follows more systematic procedures, ensuring a tighter tracking of the target values of the physiological variables to regulate, together with a reduction of the quantity of drugs administered;
• In animal surgery, where for economic reasons the surgeon many times has to play the double role of anaesthetist, an important aid will be provided by the automatic system, thereby improving the quality of anaesthesia while keeping the cost low.

According to the above motivations, the objective of this project is the development of an autonomous integrated system for the automation of anaesthesia that incorporates advanced control algorithms able to tackle the specific challenges of anaesthesia. The system consists of a multiprocessor unit that can be connected to standard anaesthesia sensors and actuators (perfusion syringes) and comprises two blocks.

One of the blocks performs the functions associated to control and comprises control algorithms embedded in a Digital Signal Processor that decides the dose of drugs to administer such that the physiological variables of the patient to control are close to the desired target values. The variables to control are the neuromuscular blockade (NB) and the depth of anaesthesia (DoA) levels. Emphasis will be placed on using the entropy measure of the EEG to evaluate the DoA and insert it a feedback loop. Furthermore, sensor fusion of different measures, such as combining the BIS and EM, to achieve more reliable signals for feedback will be considered.

The control algorithms will exploit the connection between Model Predictive Control (MPC) and Switched Multiple Model Adaptive Control (SMMAC) in order to take advantage of the specific features of the critical application considered. Besides its predictive features, MPC allows to easily incorporate constraints and feedforward from accessible disturbances, three key aspects for the control of anaesthesia. On the other side, the great variability inter and intra patient calls for the use of adaptive methods. SMMAC provides the desired adaptive features, including the possibility of controller reconfiguration, an important issue in fault tolerant control, as well as tackling signal artifacts.

The other major block consists of a supervisor, performing the functions of a Fault Detection Monitor (FDM). This block verifies the correctness of the actions undertaken by the controller block, while being able to include the adequate high level information on the patient’s state to the practitioner provided by software developed in other projects. In particular, the FDM will detect malfunctions of sensors and actuators (e. g. blocking, occurrence of outliers or unexpected behaviour), oscillatory and other poor performance behaviour of the feedback loop.

The main expected scientific contribution derives from the development of an integrated unit for automatic anaesthesia (AA) incorporating nonlinear and adaptive algorithms for control, and its demonstration on clinical cases. Due to the ethical restrictions inherent to the experimentation on human patients, the test of the final system will be mainly performed on rats subject to anaesthesia, although human patient experimentation is also considered. More specific issues include:
• Feedback based on new measurement variables such as the entropy measurement of DoA;
• New sensor fusion and sensor fault tolerant algorithms for both DoA and NB.
• Algorithms for control loop diagnosis in automatic anaesthesia and controller reconfiguration;
• On-line structure identification of the bank of controllers in SMMAC applied to automatic anaesthesia.

In order to achieve the objectives, a multidisciplinary working team has been organised comprising experts from Control and Automation (INESC-ID, FCUP), Human Anaesthesia (HGSA), Animal Anaesthesia (UTAD), Mathematical Modelling of Anaesthesia (FCUP) and Computer Architectures and Electronic Systems (INESC-ID). Most of the members from these teams have a reach past experience of collaborative research on Monitoring and Control for Anaesthesia.