To overcome the present limitations of simple threshold alarms, an automatic real-time trend analysis system is being developed to warn about slowly varying trends or short interval disturbances before the underlying condition results in a critical situation. A rule based approach is adopted to help in the detection and diagnosis of critical patient conditions. The first level performs low-level trend analysis based on median and FIR filters. A second module detects and displays symptomatic patient conditions using a decision tree of rules to process the data from the first module. The third module displays the processed data and symptoms generated to help the user interpret the advisory diagnosis.
C. Collet, A.S. Malowany