Analysis of microvascular blood flow and oxygenation: Discrimination between two haemodynamic steady states using nonlinear measures and multiscale analysis

Abstract

Objective: This study investigates the feasibility of the use of nonlinear complexity methods as a tool to identify altered microvascular function often associated with pathological conditions. We evaluate the efficacy of multiscale nonlinear complexity methods to account for the multiple timescales of processes modulating microvascular network perfusion. Methods: Microvascular blood flux (BF) and oxygenation (OXY: oxyHb, deoxyHb, totalHb and SO2%) signals were recorded simultaneously at the same site, from the skin of 15 healthy young male volunteers using combined laser Doppler fluximetry (LDF) and white light spectroscopy. Skin temperature was clamped at 33 o C prior to warming to 43 o C to generate a local thermal hyperaemia (LTH). Conventional and multiscale variants of sample entropy (SampEn) were used to quantify signal regularity and Lempel and Ziv (LZ) and effort to compress (ETC) to determine complexity. Results: SampEn showed a decrease in entropy during LTH in BF (p=0.007) and oxygenated haemoglobin (oxyHb) (p=0.029). Complexity analysis using LZ and ETC also showed a significant reduction in complexity of BF (LZ, p=0.003; ETC, p=0.002) and oxyHb (p<0.001, for both) with LTH. Multiscale complexity methods were better able to discriminate between haemodynamic states (p<0.001) than conventional ones over multiple timescales. Conclusion: Our findings show that there is a good discrimination in complexity of both BF and oxyHb signals between two haemodynamic steady states which is consistent across multiple scales. Significance: Complexity-based and multiscale-based analysis of BF and OXY signals can identify different microvascular functional states and thus has potential for clinical application in the prognosis and the diagnosis of pathophysiological conditions such as microvascular dysfunction observed in non-alcoholic fatty liver disease and type 2 diabetes.

Multi‐domain analysis of microvascular flow motion dynamics in NAFLD

“…So, the length of each time series { y τ } is equal to the length of the original signal divided by the scale factor, τ. At the coarse‐grain of 24 scales, the signal length is 1000 which we show in is sufficient for complexity analysis. The LZC was calculated for each coarse‐grained sequence as a function of the scale factor, τ, and the MLZC was evaluated.…”

“…Above the scale corresponding to the heart rate of the individual time series, the relatively periodic oscillatory influence will be diminished and the signal contains more information. To reach higher scales, longer recordings are required and we have shown that there is a lower limit on the number of samples required to achieve repeatable and accurate complexity analysis limiting the lowest frequency cutoff. We observed different scales at which the maximum separation between groups occurred with DM0‐DM1 and DM0‐CB at scale τ = 15 and DM1‐CB at τ = 10.…”

“…In previous work, we have shown that these LZC measures can be used to discriminate between two hemodynamic steady states (resting and that induced by local thermal hyperemia) and are thus a good candidate for the study presented here. In Tigno et al, using vasomotion to predict metabolic risk groups in nonhuman primates, this sequence is determined by comparing each element in the time series with the median value and replacing it with a zero if it is less than the median or a one otherwise.…”

Multiscale, multidomain analysis of microvascular flow dynamics

“…the number of unique patterns in the time series or model order; (ii) how hard it is to create or compress the information without loss; and (iii) the degree of organization or structure in LDF( t ). The most widely used measures used in LDF signal analysis are Lempel–Ziv complexity (LZC), sample entropy and effort‐to‐compress complexity (Thanaj, Chipperfield, & Clough, ). However the complexity is estimated, it provides a measure of the information content of a signal (Tigno et al., ), and for brevity, only LZC (Lempel & Ziv, ) will be considered here.…”

“…At scale factor one, the time series y 1 is the original signal, and the length of each coarse‐grained time series is equal to the original signal divided by the scale factor, τ. In previous work, we have investigated the length of signal required to obtain viable complexity measures and reported that a signal length >1000 samples is required (Thanaj et al., ), which equates to 10 min captured at 40 Hz at scale .…”