On the Validity of Neural Mass Models

Area of Science:

• Computational Neuroscience
• Neural Dynamics Modeling

Background:

• Neural mass models (NMMs) are widely used to study neural population dynamics, explaining phenomena like oscillations and synchronization.
• The accuracy of NMMs, particularly for mixed excitatory-inhibitory networks, remains a subject of investigation.

Purpose of the Study:

• To evaluate the fidelity of NMMs in representing the mean dynamics of neural populations.
• To identify the conditions and limitations under which NMMs provide accurate estimates.

Main Methods:

• Derived two population dynamics models from a network of leaky integrate-and-fire neurons, falling under Freeman's NMM category.
• Employed mean-field assumptions and time-scale separation between membrane and synaptic dynamics.
• Compared derived NMMs against the averaged dynamics of the underlying neural network.

Main Results:

• NMM accuracy is constrained by the fraction of excitatory/inhibitory neurons and network degree.
• Significant parameter ranges exist where NMMs fail to mimic network dynamics, including desynchronized and high-frequency synchronized states.
• NMMs accurately estimate mean potential dynamics primarily around the onset of low-frequency synchronization.

Conclusions:

• NMMs show potential for studying specific states like resting brain activity (e.g., from EEG) near transition regions.
• Predicting general neural network dynamics using NMMs requires careful consideration of their inherent limitations and applicable parameter ranges.