Proceedings of The Physiological Society

Europhysiology 2018 (London, UK) (2018) Proc Physiol Soc 41, PCA333

Poster Communications

Application of Texture Analysis to the spectral decomposition of Photoplethysmography vascular signals

H. Silva1,2, H. A. Ferreira3, L. Monteiro Rodrigues1,2

1. CBiOS (Research Center for Biosciences and Health Technologies), U Lusófona, School of Health Sc & Technologies, Lisboa, Portugal. 2. Pharmacol. Sciences Departm., U Lisboa, Faculty of Pharmacy, Lisboa, Portugal. 3. IBEB Institute Bioph Biomed Engineering, U Lisboa, Faculty of Sciences, Lisboa, Portugal.

Texture analysis (TA) allows the characterization of features in image patterns as mathematical descriptors. One of TA's most relevant descriptors is entropy, defined as the randomness or "disorder" of an image, showing a "coarse" or "rough" appearance, with irregular patterns. Entropy, applied to physiological signal processing, refers to the randomness of a signal over time, e.g. the competence of a biological system. Physiological signals are often complex and multiscaled, and for appropriate analysis are often decomposed into their main spectral components with mathematical tools such as the wavelet transform (WT). One of the outputs of this analysis is the 2D scalogram, a time-scale representation where the different components of the signal appear as bands of different textural patterns. Our objective was to apply texture analysis functions to assess the textural entropy (TE) of WT-derived photoplethysmography (PPG) scalograms, and compare this with the complexity index (CI), a measure of signal entropy calculated with Multiscale Entropy Analysis (MSE). PPG signals were collected from both feet of twelve healthy subjects (both sexes, 26.0 + 5 y.o.) lying supine while evoking a standard venoarteriolar reflex (VAR) test lying supine, in three phases - 10 min baseline, 10 min with one foot (test) lowered, and 10 min recovery in the initial position. All procedures complied with the Helsinki Declaration and subsequent amendments. The entropy of each PPG spectral component was quantified by two different approaches (i) as the complexity index (CI) of each component's frequency interval, and (ii) from the signal's 2D scalograms, obtained from TA. A p<0.05 was adopted. All components had significantly different CI between feet during VAR, and all CI decreased significantly in both feet, except for the cardiac component in the control foot. TA revealed a decrease in the TE of most components in both feet, but only the changes in the respiratory, NOd and NOi components were significant. All components, except the sympathetic, revealed significantly different TE between feet. MSE suggests that VAR decreases the randomness of most PPG signal components. TA seems to suggest the same, further reinforcing the conclusion that VAR induces a change in the competence of vascular regulation mechanisms. Therefore, TA seems to be an adequate instrument to further explore the oscillatory properties of vascular signals.

Where applicable, experiments conform with Society ethical requirements