‘Trust me I am a doctor’: From parsimonious models to pragmatic decision‐making and control
Mahdi Mahfouf · International Journal of Adaptive Control and Signal Processing · 2009
During the 5th IFAC World Congress which took place in Prague (Czech Republic) in 2005 Professor Rudolf Kalman delivered a keynote speech when he famously declared: ‘understand the mathematics and the rest is easy’, referring of course to solving complex problems. Surprisingly‡ perhaps, Professor Michael Athans did somehow challenge this argument by pointing out that there exist systems that do not necessarily lend themselves to mathematical descriptions, such as those associated with Management Systems, for instance, and as a consequence subjective reasoning (I suggest e.g. fuzzy logic-based reasoning) should play a more significant role in this case. It is not surprising then that a ‘hybrid’ approach is increasingly adopted in today's systems, which are becoming endemically more complex and more integrated. Biological systems are no exception since one can safely say that they always represented an ‘ideal’ test bed for hybridization. For example, to describe the effect of drugs on the body, pharmacology (pharmacokinetics and pharmacodynamics) has always played a major role in devising effective patient therapies; however, when data are not available or are sparse subjective reasoning using expert knowledge, in this case the medical staff, may prove to be vital. Although not explicitly included in this special issue, intensive care medicine is another sub-area of healthcare where a hybrid approach to describing the interactions between the critically ill patient and machine (ventilator) is necessary due to the fact that there exist processes that can be described by differential equations, such as those associated with blood–gas exchanges, but there are also other complex, nonlinear phenomena, which can perhaps be characterized conceptually via ‘subjective’ reasoning with the help of experts. If ‘lighter, faster, and cheaper’§ sums up the ever competitive market associated with the Aerospace Industry, then ‘safer’ should be the overarching message of healthcare systems. In other words, the ‘economics’ should be a consequence rather than an objective. Medical specialists in hospitals ‘optimize’ individualized treatment to make the critically ill patient traverse the first (crucial) 24 h, called the golden hours, safely then they initiate the following phases of treatment, which hopefully will allow him/her to leave the clinical environment sooner to minimize any risks associated with infection. The fact that the bed for the next patient becomes available should be more of a consequence rather than an objective in itself, which does not preclude the fact that hospitals are expensive to run: a single day reduction in therapy can save in the U.K. as much as £1.2M per year for an intensive care unit with 500+ admissions! More often than not, when I describe results of clinical trials to an audience I always begin with a message that says that this is no simulation but real-life drama. The same patient who undergoes several clinical trials, seemingly under similar conditions, may produce different results. This is typical of trials under closed-loop control conditions where the same controller can lead to different transients and steady-state behaviour. The audience indeed appreciates that one is more likely to obtain consistent behaviour if one dealt with engines rather than humans¶. Perhaps what fascinates the audience first and foremost is the multi-disciplinary nature of bioengineering research as it calls upon expertize in science and engineering. This is simply its raison d' etre. All the papers in this special issue reflect such multi-disciplinary research flavour including ‘Anaesthesia’, ‘Rehabilitation’ as well as ‘Diabetes Therapy’. Some are purely simulation-based others are a combination of simulations and real-time experiments but all authored by well-known researchers in their respective fields. In the paper by Wong et al. a protocol for the control of Type 1 diabetes with several attractive attributes, including safety, is proposed. The protocol relies on patient-monitored blood glucose measurements and insulin injections to make it effective. The results that show much improved performances were based on real patient scenarios replicated via a virtual simulator. Dumont and his co-workers argue that robustness of controllers is ‘key’ to automated drug delivery protocols being adopted as routine clinical practice. To support this, they propose a design procedure for robust control based on conventional proportional integral derivative and fractional calculus for the intravenous administration of propofol, a fast-acting anaesthetic. The paper by Nino et al. focuses on the application of De Keyser and van Cauwenberghe's Extended Predictive Self-Adaptive Control algorithm to control depth of Anaesthesia (DOA) via propofol administration. DOA was monitored via EEG using the Bispectral Index (BIS). To tackle the nonlinearities present in the system an on-line gain adaptation strategy that takes into account the drug concentration/effect relationship is also included in the control strategy. In their paper, Hunt and Allan show how to infer knowledge from man (people with neurological impairment)–machine (robotics-assisted gait device) interactions. The monitoring of patients during exercise training via performance indices is of vital importance and the linear behaviour of specific parameters as opposed to nonlinear characteristics may help with such knowledge inference. A combination of a new model structure and linear feedback control seems to induce a linear oxygen uptake by the patients, which in turn leads to a more accurate estimate of specific parameters of performance. Nunes and her co-workers describe the application of the MUSMAR adaptive control algorithm developed previously by Mosca et al. in order to adjust the level of unconsciousness, which is monitored via EEG using the BIS. The authors also argue that in order to allow the closed-loop control system to reflect the realities of the clinical environment, muscle relaxation via electromyogram measurements and analgesia therapy via the drug remifentanil need to be incorporated as secondary but nonetheless significant paths. Simanski et al. went where only a few researchers dared to go, i.e. from simulations to real-time implementation. In their paper, they introduced a multi-input/multi-output drug delivery system called the Rostock assistant system for Anaesthesia control. The system was able to successfully control unconsciousness as well as neuromuscular blockade (paralysis) simultaneously. More than 20 patients have been used in their study. In the last paper, Mahfouf et al. compared two different monitoring techniques for DOA, one referred to as (or part of) the ‘clinical gold standard’, which uses mean arterial pressure as the indicator of level of unconsciousness, whereas the second relied on a more sophisticated measurement technique that uses filtered EEG, in this case audio-evoked potentials. After the merits and limitations of both methods that are exposed through clinical trials the authors identify six overarching questions (research challenges), which, in their opinion, should pave the way to a unified approach of monitoring DOA effectively and safely. Similar to all good research, biomedical research has a good future just as Anderson predicted in the 1970s and will continue to grow in the near future as long as we all carry on looking for newer and more challenging synergies which, I believe, are just around the corner! In summary, I wish to thank the Managing Editor of this Journal, Professor Mike Grimble, for his support and encouragement towards organizing this Special Issue. Thanks should also go to the Editorial Staff at John Wiley and Sons Ltd., in particular Miss Jennifer Hillier, Miss Adele Davies and Miss Alexandria Owen, for promptly replying to my incessant emails and providing me with the necessary assistance. Last but not least, I would like to take this opportunity to thank all the contributors to this issue for their special effort in preparing their manuscript and also the reviewers for carefully scrutinising their contributions to reflect the reputation of this Journal.