Skip to main content
. 2018 Dec 7;9:5243. doi: 10.1038/s41467-018-07647-3

Fig. 9.

Fig. 9

Comparison of BCI performance between within-day decoder pairs. For each day, the decoder that was run first is on the left, and the second on the right. Asterisks represent significance level of statistical tests between decoders: *p < 0.05, **p < 0.01, ***p < 0.001 (single-sided Mann–Whitney U-test for a, b and c; chi-square test for d). On day 2–4, center-hold duration was shortened to be between 200 and 300 ms for OLE. The reference speed s0 used for decoder calibration (see Methods) was lowered to 80% for OLE on day 2–3. Task parameters of each session, including the number of trials per target, could be found in Supplementary Table 2. a Mean rate of motion during center-hold before movement onset. The marker (blue circle for OLE, orange square for Direct Regression, and yellow diamond for hANN) represents the median value across all trials, and “whiskers” represent the 25 and 75 percentile values (same in b and c). b Mean rate of motion during target-hold after movement offset. c Mean curvature of cursor trajectory between movement onset and movement offset. d Percentage of successful target-hold periods of the trials in which the target was acquired