Diagnosis and decision-making for awareness during general anaesthesia

• Published in: IEE conference publication pp. 73 - 77
• Main Authors: Backory, J.K, Abbod, M.F, Linkens, D.A
• Format: Conference Proceeding
• Language: English
• Published: London IEE 1998
• Subjects: Applied sciences; Artificial intelligence; Bioelectricity; Biological and medical control systems; Biology and medical computing; Computer science; control theory; systems; Connectionism. Neural networks; Digital signal processing; Exact sciences and technology; Function theory, analysis; Integral transforms; Knowledge engineering techniques; Knowledge engineering tools; Neural computing techniques; Optical, image and video signal processing; Patient care and treatment; Pattern recognition; Problem solving, game playing; haemodynamic variables; wavelet transforms; surgical interventions; patient monitoring; sedation; unconsciousness level; awareness; feature extraction; knowledge based systems; multiresolution wavelet analysis; fuzzy systems; biocontrol; clinical signs; drug profile; pattern classification; knowledge acquisition; general anaesthesia; diagnostic system; knowledge elicitation; closed-loop architecture; mid-latency auditory evoked potentials; closed loop systems; decision-making; neuro-fuzzy system; auditory evoked potentials; neural nets; surgery; anaesthetic depth; Fuzzy logic; Wavelet transformation; Decision making; General anesthesia; Diagnosis; Neural network; Multiresolution analysis
• ISBN: 085296708X; 9780852967089
• ISSN: 0537-9989
• DOI: 10.1049/cp:19980204

We describe the design process of a diagnostic system for monitoring the anaesthetic state of patients during surgical interventions under general anaesthesia. Mid-latency auditory evoked potentials (MLAEPs) obtained during general anaesthesia are used to design a neuro-fuzzy system for the determination of the level of unconsciousness after feature extraction using multiresolution wavelet analysis (MRWA). The neuro-fuzzy system proves to be a useful tool in eliciting knowledge for the fuzzy system: the anaesthetist's expertise is indirectly coded in the knowledge rule-base through the learning process with the training data. The anaesthetic depth of the patient, as deduced by the anaesthetist from the clinical signs and other haemodynamic variables, noted down during surgery, is subsequently used to label the MLAEP data accordingly. This anaesthetist-labelled data, used to train the neuro-fuzzy system, is able to produce a classifier that successfully interprets unseen data recorded from other patients. This system is not limited, however, to the combination of drugs used here. Indeed, the similar effects of inhalational and analgesic anaesthetic drugs on the MLAEPs demonstrate that the system could potentially be used for any anaesthetic and analgesic drug combination. We also suggest the use of a closed-loop architecture that would automatically provide the drug profile necessary to maintain the patient at a safe level of sedation.