While the laser speckle imaging (LSI) is a powerful tool for multiple biomedical applications, such as monitoring of the blood°ow, in many cases it can provide additional information when combined with spatio-temporal rhythm analysis. We demonstrate the application of Graphics Processing Units (GPU)-based rhythm analysis for the post processing of LSI data, discuss the relevant structure of GPU-based computations, test the proposed technique on surrogate 3D data, and apply this approach to kidney blood°ow autoregulation. Experiments with surrogate data demonstrate the ability of the method to extract information about oscillation patterns from noisy data, as well as to detect the moving source of the rhythm. The analysis of kidney data allow us to detect and to localize the dynamics arising from autoregulation processes at the level of individual nephrons (tubuloglomerular feedback (TGF) rhythm), as well as to distinguish between the TGF-active and the TGF-silent zones.