Effects of motion artifact on the blood oxygen saturation estimate in pulse oximetry

Oxygen saturation estimates from pulse oximeters (SpO
2
) have been shown to be unreliable in the presence of motion artifact. This may cause errors in the clinical environment if the device falsely detects normal or desaturated conditions. This paper seeks to investigate the failure modes of the standard SpO
2
calculation algorithm in the presence of motion artifact. A Texas Instruments AFE4400 evaluation module was used to collect data. The board is designed for pulse oximetry applications and allows access to the raw photoplethysmograph signals. Measurements were taken from a single subject with a finger probe. Signals were collected both while moving the instrumented hand and while moving the sensor without moving the hand. These were compared to a control signal where the subject remained motionless. Oxygen saturation was constant as verified by a Clevemed Bioradio SpO2 probe on the subject's other hand, which remained motionless for all of the measurements. The results showed a significant decrease of measured SpO
2
during motion of the hand but not during motion of the sensor. This was likely due to the probe detecting the movement of venous blood, or failure to correctly detect peaks in the PPG signals. The variability of the measured SpO
2
increased during motion of the hand and motion of the sensor, likely due to variation of the optical path length through the tissue. This work will help future development of algorithms to improve the performance of pulse oximetry in ambulatory conditions.