iNTRODUCTION: Pulse oximetry. äs a non-invasive technique, brought a major advance to patient monitoring in the 1980s, yet some of the most valuable data in the waveform signal are not used. See e.g. Murray '. this waveform can offer a valuable. noninvasive diagnostic tool for circulatory dynamics. Now the vast pool of photoplethysmographic patient data acquired from pulse-oximeters within the framework of the EC mult icentre study Pulse Oximeier Calihraior carries great potentia! for the identification of c l in ica l parameters associated with a given pulse shape. On the other hand, among various rapid developing new digi tal signal processing techniques, artificial neural network techniques show great ability on handling this di f f icu l i tasks. The purpose of this study is, therefore, to apply neural network techniques to classify the pulse shapes and determine some clinical parameters associated with the pulse shapes.