Detection and analysis of spatiotemporal patterns in brain activity

Abstract

There is growing evidence that population-level brain activity is often organized into propagating waves that are structured in both space and time. Such spatiotemporal patterns have been linked to brain function and observed across multiple recording methodologies and scales. The ability to detect and analyze these patterns is thus essential for understanding the working mechanisms of neural circuits. Here we present a mathematical and computational framework for the identification and analysis of multiple classes of wave patterns in neural population-level recordings. By drawing a conceptual link between spatiotemporal patterns found in the brain and coherent structures such as vortices found in turbulent flows, we introduce velocity vector fields to characterize neural population activity. These vector fields are calculated for both phase and amplitude of oscillatory neural signals by adapting optical flow estimation methods from the field of computer vision. Based on these velocity vector fields, we then introduce order parameters and critical point analysis to detect and characterize a diverse range of propagating wave patterns, including planar waves, sources, sinks, spiral waves, and saddle patterns. We also introduce a novel vector field decomposition method that extracts the dominant spatiotemporal structures in a recording. This enables neural data to be represented by the activity of a small number of independent spatiotemporal modes, providing an alternative to existing dimensionality reduction techniques which separate space and time components. We demonstrate the capabilities of the framework and toolbox with simulated data, local field potentials from marmoset visual cortex and optical voltage recordings from whole mouse cortex, and we show that pattern dynamics are non-random and are modulated by the presence of visual stimuli. These methods are implemented in a MATLAB toolbox, which is freely available under an open-source licensing agreement.

Dopamine waves as a mechanism for spatiotemporal credit assignment

“…The dorsal striatum exhibits graded behavioral specialty, with the DMS orchestrating goal-directed behaviors involving actionoutcome contingencies, and the DLS implicated in stimulus-response behaviors (Balleine et al, 2007;Corbit and Janak, 2010;Yin and Knowlton, 2006) . Inactivation or manipulation of dopamine in DMS degrades goal-directed planning and action due to an inability to learn whether rewards are under instrumental control (Balleine and O'doherty, 2010;Wunderlich et al, 2012) .…”

“…We characterized flow patterns in dopamine waves by adapting standard optical flow algorithms in machine vision that are adapted for imaging of fluorescence signals (Afrashteh et al, 2017;Mohajerani et al, 2013;Townsend and Gong, 2018) . Briefly, flow trajectories were computed for any two successive frames as a displacement of intensity across the pixels in time.…”

“…This method allows us to evaluate a pixel-by-pixel velocity vector fields that summarizes the direction and strength of flow at each pixel. While there are multiple methods to achieve this calculation (Afrashteh et al, 2017;Townsend and Gong, 2018) , we adapted a combined Global-Local (CGL) algorithm (Bruhn et al, 2002;Liu, 2009) that combines the Lucas-Kanade and Horn-Schunck methods. The frame-by-frame vector fields calculated using the CGL method was further processed to extract sink and source locations and also flow trajectories across multiple frames (Fig.…”

Spatiotemporal patterns of rodent hippocampal field potentials uncover spatial representations

“…A central task in computational neuroscience is to link neural representations with animal's behavior or internal cognitive state. Therefore, development of unsupervised machine learning methods is important for discovering intrinsic latent structures of high-dimensional spatiotemporal neural data, especially in the absence of behavioral measures (Cunningham and Yu, 2014;Chen et al, 2014;Linderman et al, 2016;Townsend and Gong, 2018;Chaudhuri et al, 2019). Since high-density LFP recordings produce a high degree of input correlation between neighboring channels, low-rank feature extraction or dimensionality reduction (e.g., ICA, NMF and embedding) can help exploratory data visualization and subsequent decoding analysis.…”

Cortical propagation as a biomarker for recovery after stroke

“…This is especially important when considering motor evoked activity, which is directly associated with the execution of single movements (active forelimb pulling in our case). The detailed propagation analysis we showed here would not be possible with the widely used optical flow techniques (Afrashteh et al, 2017;Townsend and Gong, 2018) which instead focus on velocity vector fields and their complex patterns (e.g. sources and sinks) but do not deal explicitly with temporal order.…”