P097 Towards Optimized tACS Protocols: Combining Experimental Data and Spiking Neural Networks with STDP
Camille Godin*1, Jean-Philippe Thivierge1,2
1School of Psychology, University of Ottawa, Ottawa, Canada. 2Brain and Mind Research Institute, University of Ottawa, Ottawa, Canada
*Email:cgodi104@uottawa.ca Introduction:Abnormal neuronal synchrony is linked to various pathological conditions, often manifesting as either excessive [1] or reduced oscillatory activity [2]. Thus, modulating brain oscillations through transcranial electric stimulation (TES) could help restore healthy activity. However, TES outcomes remain inconsistent [3], emphasizing the need for deeper understanding of its interaction with neural dynamics. Transcranial alternating current stimulation (tACS) – a form of oscillatory TES - allows for diverse waveforms, yet sinusoidal stimulation remains the predominant choice in both experimental and clinical settings. Optimizing simulation parameters could improve efficacy and reduce variability in outcomes, making TES a more reliable tool.
Methods:We modeled a Spiking Neural Network (SNN) of 1000 excitatory-inhibitory Izhikevich neurons with sparse, recurrent connectivity. We first aimed to replicate neural patterns observed in experimental data [4], where local field potential (LFP) signals were recorded from area V4 of amacaque monkey receiving sinusoidal tACS at 5, 10, 20 and 40 Hz, and SHAM. We tuned the model to match the SHAM condition(Fig 1. A), characterized by noisy delta oscillations, and then introduced external inputs to mimic experimental protocols. Next, we implemented Spike-Timing-Dependant Plasticity (STDP) on excitatory connections(Fig 1. B)and used the model to explore the effects of alternative stimulation waveforms and frequencies. Results:We performed a series of simulations using a baseline model tuned to SHAM (~3 Hz). The 40 Hz stimulation produced the largest relative increase in power at its respective frequency compared to SHAM (Fig 1. A). Both square and negative sawtooth waves consistently outperformed sinusoidal stimulation in increasing delta-gamma broadband power(Fig 1. C).When tracking the evolution of outward excitatory synaptic connections, it appears that square waves near 10 Hz induce the strongest synaptic changes between pre- and post-simulation, relative to the other tested shapes(Fig 1. C). Notably, the STDP model captured the harmonics observed in experimental data more accurately than the non-plastic model. Discussion:These findings highlight the relevance of Izhikevich-based SNNs with STDP for optimizing tACS protocols and improving their therapeutic potential. While sinusoidal waveforms remain the standard in tACS, our results suggest that square and negative sawtooth waves may be more effective at enhancing low-frequency synchronous activity in population oscillating within the delta-theta range. Additionally, square waves around 10 Hz induced stronger connectivity changes than other frequencies, aligning with experimental protocols to induce plasticity [5]. We argue that exploring diverse stimulation parameters is crucial to maximize the effectiveness of tACS for sustained network modifications and long-term effects on neural dynamics.
Figure 1. Fig 1. A) Left: SHAM condition in experiments and simulations. Right: Normalized relative power increase at four tACS frequencies. B) STDP integration in the SNN on excitatory connections, with weight distribution changes (black dot = centroid). C) Left: Changes in broadband power between baseline and inputs (no STDP). Right: Post-stimulation centroid relative to baseline, shifts across inputs. Acknowledgements C. C. Pack, P. Vieira and M. R. Krause. References 1. https://doi.org/10.1016/j.clinph.2018.11.013 2. https://doi.org/10.2147/NDT.S425506 3. https://doi.org/10.1371/journal.pbio.3001973 4. https://doi.org/10.1073/pnas.1815958116
