knowledge transfer


Adaptive and Intelligent Control for Multivariable Anaesthesia

Submitted by D A Linkens with M Mahfouf and M F Abbod,29.11.2001,IBA D


Balanced anaesthesia management in the operating theatre requires regulation of drugs to control depth of anaesthesia (unconsciousness),analgesia (pain relief) and relaxation (muscle paralysis for surgical efficiency and patient benefit).In theatre, pain relief is managed manually via open-loop assessment, since it cannot be measured in an unconscious patient. In contrast, relaxant and hypnotic drug effects can be monitored and their interactions controlled via feedback. Muscle relaxation can be monitored via neuromuscular stimulation at the hand, while unconsciousness can be inferred via blood pressure (particularly when using intravenous propofol).The challenge is to design a multivariable controller which allows for drug interactions and inter and intra-patient variability.


Both classical Generalised Predictive Control (GPC) and Self-Organising Fuzzy Logic Control (SOFLC) schemes have been investigated. The simulations have necessitated careful model elicitation (particularly for cross-coupling dynamics) for the combined use of propofol and atracurium. GPC uses an internal model for its control algorithm calculation at each time step, while SOFLC uses exploratory movements in the closed-loop system to refine its rule-base iteratively (ie. adaptively).Clinical trials have been undertaken for the GPC scheme, using a population average model for the internal calculations for control action.

Status and results

The GPC and SOFLC approaches have been evaluated in a comparative simulation study. Generally, GPC gave a smoother performance with less control actuator activity, but at the expense of having to know considerable detail of the process dynamics. The SOFLC gave more active control signals, and in several cases approached limit cycle conditions. The scaling factors required careful adjustment, while a second experimental run was necessary to refine the rule base. This is not surprising since the rule base was initially set to zero. The steady state performance of GPC tended to be superior to that of SOFLC.

Adaptivity and portability

The multivariable SOFLC scheme achieves Level 1 adaptivity in the EUNITE definition since it can cater for disturbances in the environment (mostly in the patient in this case) by self-adjustment of its rule base. Potentially, it could achieve Level 2 by commencing on-line control using a rule base for alternative drugs, but this has not been tested. More ambitiously, Level 3 can be attempted via initialisation from a zero rule base, but simulations have shown that this will require cautious and very slow adaptation i.e. long exploratory runs.

More information

Further details are in:

M Mahfouf and M F Abbod, "A comparative study of generalised predictive control (GPC) and intelligent self-organising fuzzy logic control (SOFLC) for multivariable anaesthesia", in "Intelligent Control in Biomedicine", ed. D A Linkens, Taylor and Francis,1994,pp79-132.

technical reports & case studies

executive summaries

scientific papers




search eunite with