A COMPUTER ALGORITHM FOR DIFFERENTIATING VALID FROM DISTORTED PULSE OXIMETER WAVE-FORMS IN NEONATES

Current pulse oximeter technology is fraught with a significant false alarm rate. This is mainly due to motion artifacts at the sensor site which distort the pulse waveform and render the computation of SaO(2) invalid. If the pulse waveform could be automatically recognized as ei ther normal or distorted, then only valid SaO(2) values would be displ ayed. We observed that the systolic upstroke time (Sy) of the pulse wa veform has a narrow and consistent range in normal appearing pulses. T he systolic upstroke time (Sy) is the time from the onset of systole t o the peak of the pulse waveform. Comparison of a preset range of Sy w as made against Sy obtained by computer analysis of each pulse wavefor m. Visual examination of 14,090 pulses was carried out to determine th e sensitivity and false positive rate of the algorithm. Sensitivity of computer detection of valid pulses was 92% with a positive predictive value of 92%. When used on line for continuous recording of SaO(2) in patients, this simple algorithm has the potential to decrease the fal se alarm rate of pulse oximeters and improve the accuracy of long-term SaO(2) recordings. (C) 1995 Wiley-Liss, Inc.