Mechanisms of Bottom-up and Top-down Information Flow in the Mouse Visual Cortex


  • Virág Horváth Eötvös Loránd University


Understanding population activity of sensory neurons is one of the key areas of modern systems neuroscience. Especially understanding information flow between higher and lower-level cortical areas is still in its infancy. In this work, we aim to use modern machine learning methods on publicly available neural datasets to study mechanisms of bottom-up and top-down processing in different layers of the mouse visual cortex.

Traditionally, neurons in the primary visual cortex have been characterized by artificial stimuli, such as gratings. For these, neuronal sensitivities could be described with simple features, resembling Gabor filters. However, recent studies have shown that sensitivities of visual cortical neurons markedly differ from the classical 'receptive fields' when richer stimuli like natural images are presented to the animal [1]. Technological advances allow for simultaneously recording many neurons in many different brain areas, which enables us to take one further step and study how different cortical areas and layers communicate.

The recorded neuronal data will be selected from the public database of the Allen Brain Observatory. To examine connections between populations of neurons we can use natural scenes and artificial stimuli. Changes in the collaboration between the activity of populations of neurons for natural scenes or artificial stimuli could highlight aspects of how stimuli with different complexities are processed. We can use multiple machine learning based methods to explore the connection between populations of neurons, such as reduced rank regression, canonical correlation analysis [2], or delayed latents across groups [3]. Using these approaches we also want to look at how different layers of the visual primary cortex take part in bottom-up and top-down information flow.


[1] T. D. Marks and M. J. Goard, "Stimulus-dependent representational drift in primary visual cortex," Nat Commun, vol. 12, no. 1, Art. no. 1, Aug. 2021. doi:10.1038/s41467-021-25436-3

[2] J. D. Semedo et al., "Feedforward and feedback interactions between visual cortical areas use different population activity patterns," Nat Commun, vol. 13, no. 1, Art. no. 1, Mar. 2022. doi:10.1038/s41467-022-28552-w

[3] E. Gokcen et al., "Disentangling the flow of signals between populations of neurons," Nat Comput Sci, vol. 2, no. 8, Art. no. 8, Aug. 2022. doi:10.1038/s43588-022-00282-5