Welcome to the Sched instance for CNS*2020 Online! Please read the instruction document on detailed information on CNS*2020.

P171: Neural routing: Determination of the fastest flows and fastest routes in brain networks
#### Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.

NEW!!: YouTube video with explanation: https://youtu.be/oLIrCDUWSV4

Play at a 2x playback speed. It's hilarious and I promise you can still understand what I'm saying. At 1x speed is toooo slow :).

Zoom Meeting

**Paula Sanz-Leon**, **Pierpaolo Sorrentino**, **Fabio Baselice**, **Rosaria Rucco**, **Leonardo L Gollo**, **James A Roberts**

**Background. **Large-scale brain networks (Bullmore and Sporns, 2009) are characterized by global and local functional and structural metrics (Rubinov and Sporns, 2010; Zalesky et al., 2010) that have furthered our understanding of brain function (Fornito et al., 2015). These metrics are based on the idea that information in a network flows along the shortest paths, either topological (Fornito et al., 2013) or geometrical (Seguin et al., 2018). In this work, we propose two functional network connectivity measures based on the physical concept of flow (Townsend and Gong, 2018; Sanz-Leon et al. 2020), encompassing both geometrical and temporal aspects of neural activity. We term the first measure *modal fastest flows (MFF)*, a time-averaged representation of the (fastest) flow lines revealing portions of physical space along which a particle (e.g., wave packet, information, spike) would travel at the maximal speed possible. The second measure, *fastest neural routes*, refers to a dense matrix where the weights are the average transit time a packet of information would take to travel from region ‘j’ to region ‘i’.

**Methods: generation of fastest flow lines**

We use our neural-flows toolbox (Roberts et al. 2019, Sanz-Leon et al. 2020) to derive flow fields from source-reconstructed MEG data. Fastest flow lines are then generated in 3 steps. First, we estimate flow vectors halfway between pairs of regions, transforming flow vectors into an edge property rather than a nodal property. Second, we trace a flow line starting from j, following the fastest flow to one of its nearest neighbours within a small spherical region. This process is done iteratively until reaching region i, and repeated for every possible region-pairwise combination. Flow lines are the sequences of maximal instantaneous speeds. Third, we average the values of each flow line to produce a matrix of fastest flows between pairs of regions.

**Results: modal fastest flows and fastest neural routes.**

We time-averaged the modal fastest flows (MFF), into a single matrix of conduction speeds. A comparison between functional connectivity derived from MEG timeseries and our MFF, indicates high similarity, quantified with the correlation matrix distance (cmd) (Herdin et al. 2005) -- 0 if matrices are equal, and 1 if completely different -- and in this case cmd=0.18. Paths highlighted by flow lines are not necessarily the shortest (in physical distance). Thus, we combine MFF with pairwise distance metrics to derive the fastest neural routes of information flow: the euclidean distance between pairs of regions, and the flow line lengths. Distributions of transit times are presented in Fig. 1. Our MFF matrix combined with the fibre length of structural connectome, can be used as a first approximation of heterogeneous time delays [s] in brain networks.

**References**

Bullmore and Sporns, 2009 Nat. Rev. Neurosc. 10(3), 186-198.

Fornito et al., 2015 Nat. Rev. Neurosc. 16 (3), 159-172

Fornito et al., 2013 Neuroimage 80, 426-444

Zalesky et al., 2010 Neuroimage 53 (4), 1197-1207

Roberts et al., 2019 Nat. Commun. 5;10(1):1056.

Rubinov and Sporns, 2010 Neuroimage, 52(3), 1059-1069

Townsend and Gong., 2018 PLoS Comput. Biol. 2018;14(12):e1006643.

Sanz-Leon et al. 2020 Neuroimage toolbox \- in preparation

Seguin et al., 2018 PNAS 115 (24), 6297-6302

Play at a 2x playback speed. It's hilarious and I promise you can still understand what I'm saying. At 1x speed is toooo slow :).

Zoom Meeting

We use our neural-flows toolbox (Roberts et al. 2019, Sanz-Leon et al. 2020) to derive flow fields from source-reconstructed MEG data. Fastest flow lines are then generated in 3 steps. First, we estimate flow vectors halfway between pairs of regions, transforming flow vectors into an edge property rather than a nodal property. Second, we trace a flow line starting from j, following the fastest flow to one of its nearest neighbours within a small spherical region. This process is done iteratively until reaching region i, and repeated for every possible region-pairwise combination. Flow lines are the sequences of maximal instantaneous speeds. Third, we average the values of each flow line to produce a matrix of fastest flows between pairs of regions.

We time-averaged the modal fastest flows (MFF), into a single matrix of conduction speeds. A comparison between functional connectivity derived from MEG timeseries and our MFF, indicates high similarity, quantified with the correlation matrix distance (cmd) (Herdin et al. 2005) -- 0 if matrices are equal, and 1 if completely different -- and in this case cmd=0.18. Paths highlighted by flow lines are not necessarily the shortest (in physical distance). Thus, we combine MFF with pairwise distance metrics to derive the fastest neural routes of information flow: the euclidean distance between pairs of regions, and the flow line lengths. Distributions of transit times are presented in Fig. 1. Our MFF matrix combined with the fibre length of structural connectome, can be used as a first approximation of heterogeneous time delays [s] in brain networks.

Bullmore and Sporns, 2009 Nat. Rev. Neurosc. 10(3), 186-198.

Fornito et al., 2015 Nat. Rev. Neurosc. 16 (3), 159-172

Fornito et al., 2013 Neuroimage 80, 426-444

Zalesky et al., 2010 Neuroimage 53 (4), 1197-1207

Roberts et al., 2019 Nat. Commun. 5;10(1):1056.

Rubinov and Sporns, 2010 Neuroimage, 52(3), 1059-1069

Townsend and Gong., 2018 PLoS Comput. Biol. 2018;14(12):e1006643.

Sanz-Leon et al. 2020 Neuroimage toolbox \- in preparation

Seguin et al., 2018 PNAS 115 (24), 6297-6302

Monday July 20, 2020 7:00pm - 8:00pm CEST

Slot 01

Slot 01