Monitoring cortical responses to neuromodulation on preclinical models can elucidate fundamental mechanisms of brain function. Concurrent brain stimulation and imaging is challenging, usually compromising spatiotemporal resolution, accuracy, and versatility. Here, we report on a non-invasive brain stimulation system with electronic control of neuromodulation parameters in a 9.4-T magnetic resonance imaging (MRI) environment. In the imaging scanner, multi-coil transcranial magnetic stimulation (mTMS) is delivered with a 2-coil array, and the MRI signal is measured with a radiofrequency coil. The mTMS can change the stimulus orientation with 1° resolution in a millisecond. Without physically rotating the coils, we evoked orientation-specific muscle responses after cortical stimulation on an anesthetized rat. The mTMS system was successfully implemented and tested with the small-animal MRI, showing minimal interference with B 0 and B 1 + fields and uncompromised image quality. A delay of 40 ms between the stimulation pulse and fMRI acquisition—similar or even shorter than those previously described in humans—led to artifact-free images. Concurrent electronically targeted brain stimulation and neuroimaging provides a valuable tool for exploring whole-brain network functions, endorsing more efficient treatment protocols.

We derived computationally efficient average response models of different types of cortical neurons, which are subject to external electric fields from Transcranial Magnetic Stimulation. We used 24 reconstructions of pyramidal cells (PC) from layer 2/3, 245 small, nested, and large basket cells from layer 4, and 30 PC from layer 5 with different morphologies for deriving average models. With these models, it is possible to efficiently estimate the stimulation thresholds depending on the underlying electric field distribution in the brain, without having to implement and compute complex neuron compartment models. The stimulation thresholds were determined by exposing the neurons to TMS-induced electric fields with different angles, intensities, pulse waveforms, and field decays along the somato-dendritic axis. The derived average response models were verified by reference simulations using a high-resolution realistic head model containing several million neurons. The relative errors of the estimated thresholds between the average model and the reference model ranged between -3% and 3.7% in 98% of the cases, while the computation time was only a fraction of a second compared to several weeks. Finally, we compared the model behavior to TMS experiments and observed high correspondence to the orientation sensitivity of motor evoked potentials. The derived models were compared to the classical cortical column cosine model and to simplified ball-and-stick neurons. It was shown that both models oversimplify the complex interplay between the electric field and the neurons and do not adequately represent the directional sensitivity of the different cell types. The derived models are simple to apply and only require the TMS-induced electric field in the brain as input variable. The models and code are available to the general public in open-source repositories for integration into TMS studies to estimate the expected stimulation thresholds for an improved dosing and treatment planning in the future.