Graph connection Laplacian allows for enhanced outcomes of consumer camera based photoplethysmography imaging

Abstract

Object: This work aims to introduce a novel method to mitigate the global phase deviation inherent in photoplethysmography imaging (PPGI) due to hemodynamics. Method: We model the facial vascular network captured by a consumer camera as a two-dimensional manifold, where the complex dynamics of the vascular tree lead to intricate phase variations across skin sites. Utilizing PPGI, we sample the vector field on the facial manifold encoding these intricate phase variations over different skin sites resulting from blood volume modulations. We propose leveraging the graph connection Laplacian (GCL) technique to quantify the global phase deviation, with the hypothesis that correcting this deviation can improve the quality of the PPGI signal and that the phase deviation encodes valuable anatomical and physiological information. Result: The proposed algorithm yields a higher-quality global PPGI signal by correcting the global phase deviation estimated by GCL, emphasizing waveform features such as the dicrotic notch. The perfusion map, with the global phase deviation estimated by GCL as intensity, reflects skin perfusion dynamics influenced by varying travel distances and anatomical structures. Conclusion: This algorithm enhances the quality of the global PPGI signal, facilitating the analysis of morphological parameters and showing promise for advancing PPGI applications in scientific research and clinical practice.