Haemodynamic trend detection using the median filter

Attention is given to methods to obtain clinically useful trend data from invasively monitored haemodynamic parameters in an intensive care unit. The requirements for a trend detector in the context of intensive care are defined. A median filter-based trend detector designed to perform as a preprocessor for a highly specific knowledge-based alarm system is presented. The trend detector uses the standard median filter to preprocess the data before estimating first-order trends with least squares sum line fitting. The detector´s responses to artifact input and its reliability as well as its clinical evaluation are discussed