Neural Noise

Neural GPS system and grid-cell navigation decoding

2 min read

  • Research on the brain’s ‘GPS system’ (e.g., grid cells and navigation) is relevant to neural signal processing and decoding for BCI.1
  • ScienceDaily coverage highlights computational neuroscience and neural data analysis; primary source should be confirmed for methods.1 1

Gardner updates

  • Neural signal processing and computational neuroscience work on the brain’s navigation/GPS system (e.g. grid cells) is relevant to decoding and neural data analysis. 1

Weekly enrichment (2026-07-20)

  • Grid cells encode two-dimensional allocentric location by forming hexagonal, periodic firing fields across an environment; cells sharing firing-field spacing and orientation form “modules,” and multiple modules layered together create a multi-scale representation of position.2
  • A 2024 study from Cell Press paired a deep learning model with experimental data to decode mouse neural activity, accurately determining where a mouse was located within an open environment and which direction it was facing purely from neural firing patterns.3
  • The method integrated activity across two navigation-related neuron types—head-direction cells (facing direction) and grid cells (2D location)—and, unlike many prior efforts, used experimental rather than simulated data, with probe recordings ground-truthed against video of the animal’s actual location and head position.3
  • The analysis introduced a simplicial convolutional recurrent neural network (SCRNN), a topological deep learning architecture that captures higher-order (beyond pairwise) connectivity and needs only spike counts, removing the need for similarity measurements.2
  • On head-direction data the SCRNN achieved the lowest mean absolute error versus three traditional neural-network architectures, and for grid-cell data it produced the smallest average Euclidean distance between decoded and ground-truth location.2
  • The authors describe their grid-cell decoding task as one of the first deep learning applications to decoding experimental (rather than simulated) grid-cell data.2
  • Grid-like coding is present in humans: a 2013 Drexel/UPenn/UCLA/Thomas Jefferson team used direct human brain recordings during a virtual-navigation video game to identify grid cells supporting path integration, published in Nature Neuroscience.45
  • A stated translational goal of the decoding work (a collaboration involving the US Army Research Laboratory, senior author Vasileios Maroulas) is to design machine-learning architectures that navigate unfamiliar terrain autonomously without GPS or satellite guidance.3

Footnotes

  1. https://news.google.com/rss/articles/CBMib0FVX3lxTE5lNlFadFJNSWEtdkVmNDUtZU5DVmNxTklCbTNuZnMtSnlMMzhJeXBWTTB0RUZINmVrU05yeVI2VW5aUmZhZGM3MFJQYUVUZWctOU82VXdmcUkyT3JjeGJZVmJRdjNDaFpZZTlqOXZndw?oc=5 ↩ ↩2 ↩3 ↩4

  2. https://doi.org/10.1016/j.bpj.2024.01.025 ↩ ↩2 ↩3 ↩4

  3. https://www.sciencedaily.com/releases/2024/02/240222122309.htm ↩ ↩2 ↩3

  4. https://www.sciencedaily.com/releases/2013/08/130804144401.htm ↩

  5. https://doi.org/10.1038/nn.3466 ↩

Graph View