MODELING TECHNIQUES AND THEIR APPLICATION FOR MONITORING IN HIGH DEPENDENCY ENVIRONMENTS - LEARNING-MODELS

This paper reviews the use of learning models including Bayesian class ifiers and artificial neural networks in monitoring and interpreting b iosignals. Generally learning models applied for analysis of biosignal s are 'black-box' types trained on the basis of measured signals. It i s illustrated that the training and application of learning models mor e or less follow the same sequences. The main focus is the interpretat ion of electrical signals from the brain (electroencephalogram (EEG) a nd evoked potentials (EP)). Current analysis of these signals often re veals sudden changes in the EEG or evoked potentials to be the earlies t discernible signs of inadequate perfusion of the brain. They may ref lect problems such as systemic arterial oxygen desaturation or hypoten sion arising from other body system failures during critical illness. It is suggested that these brain signals should be recorded in the cri tical care unit, and that they should form part of the annotated datab ase of biosignals established during the IMPROVE project. This would a llow for the development of new methods for on-line warning of impendi ng damage to the central nervous system, such that corrective actions could be taken before permanent damage occurred.