Reliability Assessment of Telemedicine Data by Analyzing Photoplethysmography with Deep Neural Network Technology

Photoplethysmography (PPG) is often used in telemedicine because it enables convenient measurement and provides data related to cardiopulmonary function. However PPG is difficult analyze using an automated algorithm because of its vulnerability to motion artefacts and the diversity of the waveforms according to the characteristics of individuals and diseases. Recently, as the use of telemedicine has become more frequent due to the outbreak of COVID19, the application of deep neural network (DNN) technology in the analysis of PPG and selection of reliable data has increased. In this study, PPG was analyzed using DNN techniques to reproduce the long-term potential (LTP) phenomenon in the brain. Moreover, the reliability of measuring saturation pulse oxymetry (SPO2) simultaneously was evaluated using the LTP-DNN. The LTP-DNN was able to evaluate faultless data by inspecting 58 PPG datasets, including 29 fault data, and could determine the possibility of failure in SPO2 measurement as well. Even in a moving situation, the LTP-DNN provides more accurate heartrate (HR) measurements than commercial SPO2 devices do. It can also be used to normalize the PPG waveform to identify waveform differences between individuals.