The paper describes an approach to intelligent ischaemia event detecti
on based on ECG ST-T segment analysis. ST-T trends are processed by me
ans of a Bayesian forecasting approach using the multistate Kalman fil
ter. A complete procedure, intended for use in CCU/ICU monitoring area
s, is proposed, in order to give the clinician an intelligent monitori
ng tool. The approach serves to describe trends and their changes in a
symbolic way. A novel aspect is its ability to observe certain featur
es of ST-T elevation/depression not detected by other means, and to re
ject artefacts and erroneous events. A sensivity of 89.58% and a predi
ctivity of 84.31% are obtained on selected records of the European ST-
T database. Using a restriction on event amplitude, the predictivity i
s raised to 95.55%. An ischaemia sensitivity index of 1.2 was determin
ed. The method has been shown to be a robust and practical trend analy
sis tool, and seems to be appropriate for numeric/symbolic transformat
ions in next-generation intelligent monitoring systems.