Rethinking noise floor characterisation in motor-evoked potentials.

Area of Science:

• Neuroscience
• Biophysics
• Statistical Modeling

Background:

• Motor-evoked potentials (MEPs) quantify corticospinal excitability, crucial for neuromodulatory interventions.
• Accurate MEP detection relies on distinguishing signals from background noise, yet noise floor properties are poorly understood.
• Current methods for identifying pre-stimulus motor activity are often ad hoc.

Purpose of the Study:

• To determine the probability distribution of the background noise floor in motor-evoked potentials.
• To evaluate the suitability of various statistical models for characterizing this noise.
• To improve the reliability of neurophysiological data interpretation in research and clinical settings.

Main Methods:

• Tested four probability distribution models (log-normal, gamma, GEV, normal) on experimental data from 19 healthy subjects.
• Used a mixture model of Gaussian and Laplacian distributions to simulate background electromyography (EMG) signals.
• Compared model performance in describing both experimental and simulated noise distributions.

Main Results:

• The noise floor distribution was highly skewed, contradicting assumptions of normality or log-normality.
• The Generalized Extreme Value (GEV) distribution model consistently provided the best fit for both experimental and simulated data.
• The gamma distribution model showed strong performance in simulations and was the second-best for experimental data.

Conclusions:

• GEV and gamma distribution models offer accurate characterization of the background noise floor.
• Employing these models can enhance precision in MEP detection, pre-activation identification, and variability analysis.
• Improved noise floor characterization leads to more reliable motor thresholding in clinical transcranial magnetic stimulation (TMS).