Abstract
Repetitive transcranial magnetic stimulation (rTMS) holds promise as a non-invasive therapy for the treatment of neurological disorders such as depression, schizophrenia, tinnitus, and epilepsy. Complex interdependencies between stimulus duration, frequency and intensity obscure the exact effects of rTMS stimulation on neural activity in the cortex, making evaluation of and comparison between rTMS studies difficult. To explain the influence of rTMS on neural activity (e.g. in the motor cortex), we use a neuronal network model. The results demonstrate that the model adequately explains experimentally observed short term effects of rTMS on the band power in common frequency bands used in electroencephalography (EEG). We show that the equivalent local field potential (eLFP) band power depends on stimulation intensity rather than on stimulation frequency. Additionally, our model resolves contradictions in experiments.
Introduction
Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive tool used to treat neurological conditions like depression and tinnitus by applying magnetic pulses to specific brain regions. However, the "dosage" is notoriously difficult to calibrate. Experimental studies often yield conflicting results—for instance, some researchers report that 10 Hz stimulation increases alpha-wave activity, while others find no change or even a decrease.
These discrepancies stem from a complex web of variables: stimulus duration, frequency, and intensity. Because it is difficult to isolate these factors in vivo, this study utilizes a mathematical model to simulate the "equivalent local field potential" (eLFP) and determine which parameters actually dictate the brain's response to rTMS.
Methodology
The researchers developed a random graph-based asymmetric Hopfield network with synchronous updates to simulate cortical activity.
Neuron Model and Network Architecture
The model employs McCulloch-Pitts neurons, which are simplified, threshold-based units. While less biologically detailed than Hodgkin-Huxley equations, they allow for the simulation of much larger networks (up to 16,000 neurons) without hitting computational limits. The brain area is treated as a random graph where the key variable is the average node degree ($k$)—the number of synaptic connections per neuron. An average node degree of $k=120$ was found to best match real-world EEG measurements of the motor cortex.
Simulation and Analysis
Stimulation was modeled as an external input that increases the current (depolarization) in a subset of neurons. The model's output, representing the sum of all spiking activity, was processed using a Fast-Fourier-Transform (FFT) to analyze power across standard EEG frequency bands: Delta (0.5–4 Hz), Theta (4–8 Hz), Alpha (8–12 Hz), Beta (12–24 Hz), and Gamma (24–48 Hz).
Key Insights and Results
Intensity Over Frequency
The most striking insight is that stimulation frequency (above 0.5 Hz) has almost no significant effect on the resulting power spectra. Instead, the intensity of the pulse is the dominant factor. This finding helps resolve long-standing contradictions in clinical data. For example, the model explains why previous alpha-band studies disagreed: because alpha activity decreases linearly with increasing intensity, different studies using different intensities would naturally report different results.
Frequency Band Dynamics
- Alpha and Beta Stability: The alpha band showed a strong linear decrease ($r = -1, p < 0.05$) as stimulation intensity increased. In contrast, the beta band remained remarkably stable regardless of the protocol.
- Gamma Growth: Gamma power showed a near-perfect linear growth ($r = 0.99, p < 0.05$) relative to stimulation intensity, suggesting gamma activity is a robust marker of rTMS-induced excitability.
- Intensity Thresholds: The model demonstrated that low-intensity stimulation (4–8 units) decreased delta power, while high-intensity (10 units) increased it, highlighting a non-linear "flip" in behavior based on dosage.
Scale Independence
The researchers found that the total number of neurons (tested from 1,000 to 16,000) didn't change the results, but the density of connections did. This implies that rTMS effects can be predicted by looking at local connectivity patterns rather than the total volume of the stimulated area.
Limitations and Future Work
The model focuses on immediate effects (the first 1000ms) and does not yet account for long-term synaptic plasticity or the spatial drop-off of magnetic field strength seen in real-world TMS coils. Future work aims to extend this model to depict pathological brain activity (like that seen in depression or epilepsy) to predict how rTMS might "reset" abnormal neural oscillations toward a healthy state.
Summary updated