Exploring the Unsupervised Detection of Multi-Neuronal Spiking patterns and the Role of Precise Timing in Neuroscientific Studies

vrijdag 31 mei 2024, 10:30
B.A. Sotomayor Gomez
prof. dr. M.A. Vinck, prof. dr. F.P. Battaglia

Understanding how neurons communicate is a key challenge in neuroscience. The complex communication in neural circuits is crucial for complex brain functions. Recent technologies like Neuropixels and SiNAPS probes allow us to record hundreds of channels simultaneously, providing new opportunities for the exploration of the neural code. The traditional rate-code theory contrasts with spike-based theory, emphasizing the temporal structure of spike trains, and current spike-based methods still exhibit challenges when applied to large ensembles of neurons. However, current spike-based methods combine both temporal and rate information, still exhibit sensitivity to rate information, or their computation becomes intractable when applied to large ensembles of neurons.

In this thesis, I investigate the spike-based view and its comparison to rate-based methods. I introduce SpikeShip, an unsupervised dissimilarity metric based on optimal transport theory, to precisely quantify temporal information from extensive neuron ensembles. I compare rate- and spike-based methods using recordings of visual areas, as well as spontaneous activity, in mice. Furthermore, I show the stable encoding of natural movies based on multi-neuron temporal spiking patterns and compare it with the rate-based schema. To facilitate the study and exploration of the neural code, I will also present an open-source Python toolbox designed to handle large neural population datasets. This toolbox offers comprehensive functionality for extracting high-dimensional neural patterns and employs a multi-core parallelization framework for efficient computation. 

As we move forward, by continuing to refine our methods and adapting to the evolving landscape of research, this work contributes to the ongoing conversation in neuroscience about the relevance of spike-based theory. Finally, this thesis also exemplifies the interdependent nature of scientific progress and the tools and methodologies we utilize to explore the dynamical nature of the brain.