Imaging of neural oscillations with embedded inferential and group prevalence statistics
Fig 7
Simulation results comparing sensitivity of iES and standard approach.
a) Examples of simulated time-series that follow a 1/f spectral distribution (grey trace) or target a pre-specified fnarrow, which is the ratio between narrow-band and broadband power (blue traces). b) Simulation setup: Two sources of interest in blue targeting pre-specified fnarrow (blue traces) are embedded in background brain noise composed of 1/f signals evenly distributed across 66 locations. c) Metric of detection probability: We quantified the probability that the two sources of interest were detected in a source map by using a range of different thresholds: the two sources were detected, if they were contained in two separate clusters after thresholding. Here we show 4 different thresholds in two simulation scenarios using the standard imaging approach. In the first scenario, sources were detected with 2 out of 4 (detection probability: 0.5) threshold values. In the second scenario, sources were detected only with 1 out 4 (detection probability: 0.25) threshold values. This configuration illustrates the issue of concurrent sources with different strengths on the detection of separate clusters of activity. d) Comparison of methods: the maps from each simulation run were thresholded using 50 different values to estimate a detection probability as in c). Since the range of data values for both MNE and iES were different, we normalized the detection probability by the maximum value obtained in each method. Thus we did not compare the absolute detection probability between the two methods, but rather how it varied with respect to the difference in fnarrow, between the sources of interest.