Segmentation of Renal Perfusion Signals From Laser Speckle Imaging Into Clusters With Phase Synchronized Dynamics

Abstract

Renal perfusion signals contain dynamics arising from the renal autoregulation feedback mechanisms as the contraction and dilation of vessels alter flow patterns. We can capture the time-varying dynamics at points across the renal surface using laser speckle imaging. We segment an imaged area of the renal cortex into clusters with phase synchronized dynamics. Our approach first uses phase coherence with a surrogate data derived threshold to identify synchronized pixel pairs. Non-negative matrix factorization is then applied to segment phase coherence estimates into phase synchronized regions. The method is applied to laser speckle imaging of the renal cortex of anaesthetized rats to identify regions on the renal surface with phase synchronized myogenic activity. In three out of six animals imaged after bolus infusion of N(ω)-nitro-l-arginine methyl ester (NAM), the renal surfaces are segmented into clusters with high phase coherence. No more than two clusters were identified during control period for any animal. In the remaining three animals, a strong myogenic signal could not be detected in surface perfusion during control or NAM. This method can be used to identify synchronization in renal autoregulation dynamics across the renal surface.

Laser speckle contrast imaging reveals large-scale synchronization of cortical autoregulation dynamics influenced by nitric oxide

“…Previous studies showed that record lengths of 30 -50 cycles (300 s) are required to probe MR synchronization (36), and pilot studies showed the same 30-to 50-cycle (1,500 s) requirement for probing TGF synchronization. BP, RBF, and LSCI signals were acquired during spontaneous BP fluctuation before and after injection of L-NAME (Sigma, 10 mg/kg iv) in 10 rats.…”

“…Synchronization of oscillators can be addressed by assessment of phase locking or of frequency entrainment (31); however, phase locking is the standard definition of synchronization (1,31). Thus determination of synchronization on the surface of the kidney used methods described in detail elsewhere (36) that are based on assessment of the stability in time of phase relationships among pixels. Briefly, each image was filtered with a Gaussian spatial filter (8-pixel width) and then down-sampled by a factor of 4, yielding a 27 ϫ 37-pixel image for each frame with average pixel size of 135 ϫ 135 m (36, 37).…”

“…PC for TGF was computed over the entire 1,500-s record, while PC for MR was computed over 300-s segments with 50% overlap. We did this because the threshold for significance decreases as the number of cycles increases in a data set, so a similar number of cycles was used to analyze MR and TGF (36).…”

“…While there are data in the literature that are consistent with wide-scale synchronization of MR, they do not demonstrate the phenomenon (13). Recently, laser speckle contrast imaging (LSCI) has been shown to capture the dynamics of renal autoregulation at the cortical surface (21,36,37). Holstein-Rathlou et al (21) identified strong synchronization of TGF dynamics in small groups of efferent arterioles (up to n ϭ 7) and interpreted the data in terms of local synchronization.…”

“…Scully et al (36) showed that renal surface perfusion can be segmented into clusters based upon synchronized MR dynamics and that these clusters are surprisingly large. The importance of addressing MR dynamics is that failure of myogenic autoregulation and development of renal failure have been closely linked both functionally and genetically (8,28), thus providing strong evidence that the MR is both necessary and sufficient for effective autoregulation.…”