Extraction of surrogate respiratory activity from pulse oximeter signals using SSA

Patient's condition in operation theaters (OTs), postoperative ICUs and critical care units (CCUs) is continuously monitored for the vital parameters such as pulse rate, oxygen saturation, breathing rate, blood pressure and others. For monitoring patients, not only in these units as well as in special situations such as cardiopulmonary and sleep disorder investigations, the breath related activity of the patient can be extracted from other physiological signals having respiratory influence such as pulse oximeter's photoplethysmographic (PPG) signals. In such situations accurate extraction of respiratory information from the PPG signal is essential to make it an alternative to all the cumbersome direct recording methods and viable under clinical settings. In this paper, we present a simple yet effective method, based on singular spectrum analysis (SSA), for extraction of respiratory activity from PPG signals. The method is applied on, PPG data collected from 12 different subjects in the laboratory with the help of an indigenously constructed prototype circuit. Each such data set consists, simultaneously recorded PPG and reference respiratory signals. The extracted respiratory signals are compared with the reference respiratory signals. Statistical parameters such as relative correlation co-efficient (RCC) in time domain as well as magnitude squared coherence (MSC) in frequency domain are used for performance evaluation along with error analysis using the accuracy rate (AcR) and normalized mean square error (NRMSE). Experimental results have shown a good acceptance for the extracted signal when compared to the originally recorded respiratory signal. The proposed technique could become an efficient approach for extraction of respiratory activity from PPG signals, avoiding usage of additional specialized sensor for recording of respiratory activity of the patient.