Fig. 2: Assessment of neurofeedback regulation performance.

A An exemplary run illustrates how the feedback score presented the feedback condition is calculated (vertical dashed lines delineate baseline and feedback conditions). Computing the feedback score of a time point during the feedback condition is performed by subtracting the median NF-signal of the previous baseline condition (in orange) to the current NF-signal and normalizing by the maximum dynamic range over a trial moving window (dashed horizontal lines). B Non-parametric amplitude adjusted surrogate data procedure to identify runs with high regulation. The NF-signal of a run is extracted and transformed into Fourier space. Surrogates with identical temporal properties are generated by randomizing the phase and doing an inverse transformation back into time domain (100,000 times). By virtually computing the average feedback score of each surrogate, a null distribution of average feedback scores of the surrogates is obtained, and the observed feedback score can then be compared against this null distribution based on an 84.1% percentile threshold. C Examples of null distributions of different runs are shown on the left. On the right, it is shown how the probability of obtaining the observed number of high regulation runs (52) or more, was obtained. I.e., by evaluating the area under the binomial distribution curve with parameters X ~ Bin(176, 1–0.841). D The number of high regulation runs is shown on the left, in function of the choice of threshold. On the right the probability of obtaining such number of high regulation runs (or more) is shown, taking into account the different percentile thresholds. Independently of the chosen threshold, regulation was successful in such a manner that the probability of obtaining the observed numbers of high regulation runs is always extremely low, even when considering null hypotheses with extremely stringent percentile thresholds.