A Comparison of Truncation Thresholds and Otsu-Based Thresholds in Predicting Behavior from Spike-Band Extracellular Neural Recordings

In extracellular neural recording data analysis, the quest for optimal amplitude thresholds for robust information extraction stands as a critical endeavor. Our study delves into two distinct thresholding methodologies: Truncation Thresholds and Otsu-based thresholds. The mean and the standard deviation of the subthreshold data segmented by the truncation thresholds are known to be good predictors of behavioral variables. On the other hand, Otsu-based thresholds have been shown to estimate subthreshold data's standard deviation more accurately in simulated data. The present study applies both methods to a real data set and reveals that the mean and the standard deviation of the subthreshold data segmented by either method are equivalent predictors of behavioral variables. We systematically gauge the prediction accuracy of the two methods and assess their computational efficiency. In light of computational considerations and real-time applicability, our research contributes to the evolution of amplitude thresholding techniques, thereby promoting their refinement for behavior prediction from neural recordings.