A technique based on the concept of a 'fuzzy trend template' has been devel
oped to identify characteristic patterns in multiple time-series. The metho
d has its foundation in fuzzy logic and allows for the intuitive and transp
arent description of 'templates', which preserve nuances of vagueness, temp
oral relationships and quantitative descriptors. Evaluation of fuzzy trend
templates can provide both belief and plausibility information for use in d
iagnostic applications. The technique has been applied to the diagnosis of
specific problems in anaesthesia and has demonstrated sensitivity and speci
ficity of 95 and 65%. respectively. Evaluation of fuzzy trend templates is,
computationally, relatively efficient and has allowed a real-time implemen
tation. The technique has the potential to be useful in any domain that req
uires temporal pattern recognition based on linguistic rules. (C) 1999 Else
vier Science B.V. All rights reserved.