Outlier Detection and Removal Signal Processing for Wearable Transcutaneous Oxygen Sensor

Respiratory diseases impact millions of people worldwide. In the smart and connected health era, there is a pressing need for reliable medical devices for remote monitoring of patients with respiratory disorders and patients with chronic diseases in general. For respiratory monitoring, there exist widely-used miniaturized photoplethysmography devices to measure peripheral blood oxygen saturation. More recently, transcutaneous oxygen monitors have been designed in a small form factor to be used as wearable devices to enable remote and continuous measurement of transcutaneous oxygen. In this study, we focus on processing the raw data collected from a noninvasive transcutaneous blood gas monitor that is in the development phase. Post-processing should be applied to remove the outliers in the measurements and smooth out the measurement noise. Various filters to remove outliers were tested in MATLAB, including the Hampel, rmoutliers, filloutliers, and moving average filters. The Hampel filter yielded the best results while maintaining the integrity of the data. The fillmissing function in MATLAB using the previous or linear interpolation methods is used to replace outlier data. The comparison of the raw data and the result of the data processing is presented.