Optimizing real-time phase detection in diverse rhythmic biological signals for phase-specific neuromodulation

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Closed-loop, phase-specific neurostimulation modulates brain activity using real-time phase estimation.
• Fast Fourier Transform (FFT)-based algorithms are common for phase detection due to efficiency and robustness.
• Optimization of FFT-based algorithms requires understanding their dependence on signal spectral properties.

Purpose of the Study:

• To evaluate the performance of different phase detection algorithms under varying signal properties.
• To identify key parameters influencing algorithm accuracy for real-time neurostimulation.
• To provide a method for optimizing FFT-based phase detection algorithms.

Main Methods:

• Offline simulations of three phase detection algorithms (endpoint-corrected Hilbert Transform, Hilbert Transform, phase mapping).
• Evaluation on diverse biological signals: rodent hippocampal theta, human EEG alpha, and human essential tremor.
• Validation of simulation predictions using real-time hippocampal theta phase detection in freely behaving rats.

Main Results:

• Algorithm performance is more sensitive to signal amplitude and frequency variations than signal-to-noise ratio.
• The size of the data window for phase estimation critically impacts FFT-based algorithm performance.
• Optimal data window size for FFT algorithms corresponds to the oscillation's period.

Conclusions:

• Signal properties like amplitude and frequency variation significantly affect phase detection algorithm performance.
• Optimizing the data window size, matching the oscillation period, is key for enhancing FFT-based phase detection.
• This study offers a practical approach to optimize algorithms for precise, phase-specific neurostimulation.