The relationship between frequency content and representational dynamics in the decoding of neurophysiological data.

Area of Science:

• Neuroscience
• Signal Processing
• Computational Neuroscience

Background:

• Decoding neurophysiological data is crucial for understanding brain function.
• A key relationship between neural signal frequency spectra and decoding accuracy is often overlooked.
• Instantaneous decoding paradigms can introduce frequency-related artifacts.

Purpose of the Study:

• To elucidate the relationship between neural signal frequency spectra and decoding accuracy timecourses.
• To propose methods for mitigating artifacts and improving the interpretation of decoding results.
• To enhance the reliability and stability of decoding metrics in neurophysiological data analysis.

Main Methods:

• Analysis of frequency spectra in stimulus-evoked neurophysiological data.
• Simulation of instantaneous signal decoding paradigms to observe frequency translation.
• Development and application of decoding paradigms incorporating signal magnitude and local gradient.
• Validation on a publicly available magnetoencephalography (MEG) dataset.

Main Results:

• Sinusoidal components in neural signals are translated to double their original frequency in decoding accuracy timecourses.
• Recommended low-pass filtering with a cut-off at one-quarter of the sampling rate to reduce aliasing artifacts.
• Decoding paradigms using signal magnitude and gradient yield higher, more stable decoding accuracy.
• The proposed methods resolve previous technical and interpretational challenges in decoding MEG data.

Conclusions:

• Awareness of frequency-dependent decoding artifacts is essential for accurate interpretation of neurophysiological data.
• Implementing specific filtering and advanced decoding features improves the robustness and interpretability of brain signal decoding.
• This work provides a framework for more reliable analysis of neural data, linking decoding performance to underlying signal characteristics.