Emergence of a Stable Cortical Map for Neuroprosthetic Control
(A) Tuning properties of a single unit in BC across multiple days. The top panel shows the spiking activity of a neuron during BC (n = 26 trials per target). The rasters are arranged to reflect the eight targets in the center-out task (target angles shown in red). Also shown are 200 randomly selected waveforms from two separate sessions (days 9 and 19). The inset shows overlapping tuning curves for each daily session for 10 consecutive days of BC. The shown voltage (V) is the peak-to-peak amplitude for both the shown waveforms. (B) Tuning properties of another unit in BC for ten consecutive days. The panel is arranged similarly to (A). (C) The three color maps to the left illustrate ensemble tuning in BC for days 3, 14, and 18 (specific units from [A] and [B] are labeled accordingly). The units were sorted on day 3 with respect to preferred direction. On the right is a color map of pairwise correlations of ensemble tuning (i.e., map) for each BC session. Warm colors represent a higher level of correlation. (D) Comparison of the learning rate with changes in ensemble tuning for two monkeys. The black solid line reproduces the learning rate from Figure 2. The red solid line represents the average correlation between a daily map and all other ensemble maps (directional tuning was assessed with 2 s of activity relative to intended target. The dotted red line shows the same relationship for directional tuning assessed with a 0.2-s window. The red line with superimposed red dots illustrates the relationship for directional tuning relative to actual cursor movements.