Poster D78, Monday, March 26, 8:00-10:00 am, Exhibit Hall C
Filtering improves skin-conductance response measures in the fMRI environment
Anthony Privratsky1, Keith Bush1, Josh Cisler2; 1University of Arkansas for Medical Sciences, 2University of Wisconsin-Madison
Skin-conductance response (SCR) data is frequently sought as a measure of sympathetic arousal in psychological research for its ease of implementation and lack of suitable alternative measures. Yet, SCR data is prone to corruption, particularly in the fMRI environment. Artifacts may take the form of slow, nonlinear drift or rapid spikes. We provide evidence that 1) difficulties in SCR data quality control are frequent causes of data corruption and exclusion, 2) researchers frequently report insufficient filtering methods and low-frequency (high-pass) filtering is typically not discussed, 3) low and high-pass filtering remove the most significant sources of noise, and 4) high-pass filtering is necessary for accurate SCR convolutional model-based response measures and increases the sensitivity of traditional peak-scoring measures. Using SCR datasets from 44 women undergoing a two-part fear conditioning and extinction task, we demonstrate the effect of high-pass filtering on derived response measures for traditional peak-scoring analysis as well as model-based regression analysis. We regressed amplitude scores onto model-based regression coefficients within participants and found that mean variance explained increased from 32% to 37% in 87 datasets (session 1, p = 9x10-4) and from 50% to 61% in 88 datasets (session 2, p = 0.003), respectively, with filtering. We recommend implementation of a SCR data processing pipeline that includes high-pass filtering and suggest that standardization of this technique will minimize lost research productivity due to difficulties in data quality control, decrease sampling bias by reducing the need to exclude datasets, and increase the validity of SCR response measures.
Topic Area: METHODS: Electrophysiology