Isolating single cycles of neural oscillations in population spiking

Neural oscillations are prominent features of brain activity, characterized by frequency-specific power changes in electroencephalograms (EEG) and local field potentials (LFP). These oscillations also appear as rhythmic coherence across brain regions. While the identification of oscillations has primarily relied on EEG and LFP, they are also present in neuronal spiking. However, several questions remain unanswered: How do spiking oscillations relate to field potential oscillations? How are spiking oscillations correlated across brain regions? And how are they connected to other physiological and behavioral measures. In this study, we explore the potential to detect individual cycles of neural rhythms solely through the spiking activity of neurons, leveraging recent advances in the high-density recording of large neuronal populations within local networks. The pooled spiking rate of many neurons within a local population reflects shared variation in the membrane potential of nearby neurons, allowing us to identify cyclic patterns. To achieve this, we utilize a Long Short Term Memory (LSTM) network, pre-trained on synthetic data, to effectively isolate and align individual cycles of neural oscillations in the spiking of a densely recorded population of neurons. We applied this approach to robustly isolate specific neural cycles across various brain regions in mice, covering a broad range of timescales, from gamma rhythms to ultra-slow rhythms lasting up to hundreds of seconds. These ultra-slow rhythms, often underrepresented in the LFP, were also detected in behavioral measures of arousal, such as pupil size and mouse facial motion. Interestingly, these rhythms showed delayed coherence with corresponding rhythms in the population spiking activity. Using these isolated neural cycles, we addressed two key questions: 1) How can we account for biological variation in neural signal transmission timing across trials during the sensory stimulation experiments? By isolating gamma cycles driven by sensory input, we achieved a more accurate trial alignment in the sensory stimulation experiments conducted in the primary visual cortex (V1) of mice. This alignment accounts for biological variability in sensory signal transmission times from the retina to V1 across trials, enabling a clearer understanding of neural dynamics in response to sensory stimuli. 2) How do spiking correlations across brain regions vary by timescale? We used the distinct spiking cycles in simultaneously recorded brain regions to examine the correlation of spiking across brain regions, separately for different timescales. Our findings revealed that the delays in population spiking between brain regions vary depending on the brain regions involved and the timescale of the oscillations. This work demonstrates the utility of population spiking activity for isolating neural rhythms, providing insights into oscillatory dynamics across brain regions and their relationship to physiological and behavioral measures.