Pipeline to analyse Photoplethysmogram (PPG) Signal Quality

Photoplethysmogram (PPG) has several potential applications; however, its computerised analysis is limited due to the presence of artifacts in the recordings. To overcome this, we present the development and validation of a preprocessing pipeline designed to detect and remove segments with noise and other artifacts in the PPG recordings. In the proposed method, PPG signals were segmented into 30-second intervals, followed by bandpass filtering to remove baseline drift and high-frequency noise. Template for noise-free signal was generated automatically from these segments and this was correlated with all the segments to detect the good quality segments. There were total 14,400 segments which were manually examined, and 11,476 were labeled as noise-free, with balance 2,924 being unsuitable for further analysis. This was considered as the ground truth for evaluating the algorithm. The results show that the algorithm achieved an accuracy of 98.03%, precision of 98.58%, sensitivity of 98.95%, F1 score of 98.76% and specificity of 94.43%. This method has the potential of computerized detection and removal of segments that have artifacts, leading to improving the quality of the remaining segments that are suitable for computerized analysis.Clinical relevance- This work will result in improving the quality of PPG recording during sleep studies, making it suitable for computerised analysis. This will support the use of PPG for sleep assessment.