Spike-to-excite: photosensitive seizures in biologically-realistic spiking neural networks
Keywords: photosensitive epilepsy, spiking neural network, V1, prediction, deep brain stimulation
TL;DR: We developed a spiking neural network model of photosensitive epilepsy that exhibits seizure-like activity in response to harmful visual stimuli, offering new insights into the condition's mechanisms and potential treatments.
Abstract: Photosensitive Epilepsy (PE) is a neurological disorder characterized by seizures triggered by harmful visual stimuli, such as flashing lights and high-contrast patterns. The mechanisms underlying PE remain poorly understood, and to date, no computational model has captured the phenomena associated with this condition. Biologically detailed spiking networks trained for efficient prediction of natural scenes have been shown to capture V1-like characteristics. Here, we show that these models display seizure-like activity in response to harmful stimuli while retaining healthy responses to non-provocative stimuli when post-synaptic inhibitory connections are weakened. Notably, our adapted model resembles the motion tuning and contrast gain responses of excitatory V1 neurons in mice with optogenetically reduced inhibitory activity. We offer testable predictions underlying the pathophysiology of PE by exploring how reduced inhibition leads to seizure-like activity. Finally, we show that artificially injecting pulsating input current into the model units prevents seizure-like activity and restores baseline function. In summary, we present a model of PE that offers new insights to understand and treat this condition.
Primary Area: applications to neuroscience & cognitive science
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 10718
Loading