Patient illness classification using time-frequency features derived from the photoplethysmogram

We report on preliminary results from a study in which statistical features derived from the time-frequency decomposition of the plethysmographic waveform were used as input to a classifier of patient illness severity. It is shown how features derived from a wavelet decomposition can be used to identify ´ill´ patients from a mixed ´ill´/´healthy´ data set. Further, an attempt to stratify according to patient illness severity is described.