Abstract
BACKGROUND: Transcranial magnetic stimulation (TMS) is a widely used non-invasive brain stimulation technique, but the neural circuits activated by TMS remain poorly understood. Previous modeling approaches have been limited to either simplified point-neuron networks or isolated single-cell models that lack synaptic connectivity. OBJECTIVE: To develop and validate a multiscale cortical circuit model that integrates morphologically-realistic neurons with accurate TMS-induced electric field distributions and to investigate mechanisms underlying cortical responses to stimulation. METHODS: We constructed a network model of a cortical column comprising 10,000 neurons across layers 2/3, 5, and 6 with over 10 million synaptic connections. The model incorporated thalamic and non-specific corticocortical inputs to generate physiological firing rates. TMS-induced electric fields were calculated using finite element modeling and coupled to individual neurons through the extracellular mechanism. We validated model predictions against experimental recordings of TMS-evoked local field potentials (LFPs) and multiunit activity. RESULTS: The model reproduced key experimental observations including the dose-dependent N50 LFP component and multiphasic multi-unit responses, consisting of an (excitatory) increase followed by a (inhibitory) decrease in firing rates. The early excitatory response exhibited dual-peak dynamics reflecting distinct contributions from directly and indirectly activated neuronal populations, and the subsequent inhibitory phase reflected activation of feedback GABAergic circuits through both GABA(A) and GABA(B) conductances. Spatial analysis across 30 cortical columns distributed across the precentral gyrus revealed orientation-dependent evoked responses. CONCLUSION: This validated multiscale model provides mechanistic insights into TMS-evoked cortical dynamics, demonstrating how direct neuronal activation cascades through synaptic networks to generate characteristic population responses. The framework establishes a computational platform for optimizing stimulation protocols in research and clinical applications.