ICA-Based Improved DTCWT Technique for MA Reduction in PPG Signals With Restored Respiratory Information

Abstract

In addition to estimation of arterial blood oxygen saturation (SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> ), pulse oximeter's photoplethysmographic (PPG) signals can be well utilized for extracting the vital respiratory information. The motion artifacts (MA) in PPGs not only make SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> estimations unreliable and inaccurate but also make it difficult to extract respiratory information. Addressing this issue, for the first time, we propose a novel approach called “ICA-based improved dual-tree complex wavelet transform (I <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> DTCWT)” technique, for efficient reduction of MAs leaving the respiratory information undisturbed. The method makes use of source separation ability of independent component analysis (ICA) along with computationally efficient modified DTCWT processing. A prototype pulse oximeter was developed and performance analysis of DTCWT, modified DTCWT and I <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> DTCWT processing methods was carried out using PPG data recorded with intentionally created MAs (horizontal MA, vertical MA, and bending MA). Experimental results demonstrated the efficiency of DTCWT processing methods in restoring PPG morphology and proved that there is a significant improvement guaranteed in reducing MAs with the presented methods. Statistical performance is evaluated in terms of measures like signal-to-noise ratio, normalized root mean square error, and correlation analysis with correlation co-efficient measure. The I <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> DTCWT outperformed other DTCWT processing methods in respect of MA reduction and the computed spectra revealed that safe extraction of respiratory information is guaranteed from these MA reduced PPGs. The proposed method is also validated by comparing with the well established signal extraction technology of MASIMO pulse oximeters, for which the discrete saturation transform (DST) is the key element. The %SpO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> estimations from processed PPGs by the proposed method closely followed the estimations based on DST and were very close to that of clean sections of PPG. In addition, the proposed method resulted in less computation cost compared to the MASIMO SET. Digital volume pulse waveform contour analysis is also performed on MA reduced PPGs to validate PPG morphology and the conventional parameters are calculated for assessing the arterial stiffness.