Fuzzy Measure Theory
A data mining framework for time series estimation
Journal of Biomedical Informatics
| Hi-index | 0.00 |
The task of administering anaesthesia requires the clinician to be vigilant for long periods of time to detect the onset of adverse conditions. Large amounts of data must be analysed in real-time and, if a problem is detected, it must be diagnosed as a matter of urgency, this being done while other management protocols are being carried out. For these reasons it would be of benefit if automated decision support could be provided for anaesthesia monitoring, to lighten the cognitive load on the anaesthetist. The Sentinel anaesthesia monitor has been developed with this objective in mind. It uses a fuzzy time-domain pattern matching technique, termed fuzzy trend templates, to detect vaguely specified patterns in multiple physiological data streams. These patterns are representative of symptoms associated with undesirable patient states. The system is capable of detecting trends and states such as 'significant rise' and 'high', and associating vague duration and temporal intervals with individual trends. Fuzzy trend templates have proven to be quite intuitive to specify, given linguistic (anaesthetists') knowledge about the problem domain. Sentinel's implementation of fuzzy trend templates also uses an extension to fuzzy logic based on the theory of evidence, to handle situations where desired information is not available, for example, when sensors are not being used. In off-line testing, Sentinel has achieved sensitivity and specificity of above 90% in the diagnosis of seven common or serious conditions that can arise during anaesthesia.