Average of Baseline Autocorrelation Function is a Leading Indicator of Neural Stimulation Data Quality

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Experimental neural stimulation data quality varies, impacting research efficiency and statistical power.
• A need exists for early, cost-effective indicators to predict and potentially improve data quality during sessions.

Purpose of the Study:

• To introduce and validate a leading indicator for neural data quality.
• To assess the indicator's predictive power for neural response forecasting.

Main Methods:

• Calculated the average autocorrelation function over multiple lags from ≈5 minutes of baseline neural data.
• Utilized long short-term memory (LSTM) networks and temporal basis function models (TBFMs) to forecast neural responses.

Main Results:

• The autocorrelation-based indicator was highly correlated with the ability to forecast neural responses.
• The indicator serves as a proxy for signal noise level and is predictive of data quality.

Conclusions:

• The developed indicator is a fast, inexpensive, and actionable tool for assessing neural data quality early in experimental sessions.
• This method can potentially prevent the waste of valuable research time and effort by enabling early intervention.