MAKING SENSE OUT OF COMPLEXITY: METHODS FOR PROCESSING WAVEFORMS AND IMAGES.

Computers are used at many levels of machine intelligence within signal and image processing systems. At the lowest level, digital computation may be used simply to clean up or enhance the measured signal and display it in a vivid and meaningful form. At the next level, the computer may be programmed to extract from these signals predetermined features which preserve the medical significance while eliminating redundancy and irrelevant detail. These features may be of immediate interest to the medical user of the system, such as statistical or structural parameters of the signal (frequency band description of the EEG, cardiac output from thermal dilution signal), in which case their displayed values complete the evaluation cycle. At the highest level of machine intelligence, the features may be abstract attributes of the signals and images, which serve as input to classification and pattern analysis procedures. Here the computer derives conclusions about the signal, based on the data itself and a knowledge base which puts the data into context.