Nature Neuroscience
6, 1253 - 1254 (2003)
Published online: 21 November 2003; | doi:10.1038/nn1158
Inference of hand movements from local field potentials in monkey motor cortexCarsten Mehring, Jörn Rickert, Eilon Vaadia, Simone Cardoso de Oliveira, Ad Aertsen
& Stefan RotterSupplementary Fig. 1. (jpg 16K)
Temporal evolution of tuning strength. Colors depict the different types of signals: LFPs (green), SUAs which were checked for stability (red) and MUAs (blue). Tuning strength was calculated as the signal-to-noise ratio of the tuning curves in windows of 50 ms width from the unsmoothed signals. Supplementary Fig. 2. (jpg 11K)
Decoding power and accuracy of trajectory prediction for different algorithms. We used the neuronal signals (LFP, SUA, MUA) that were recorded simultaneously by eight electrodes. The graphs depict averages over 10 recording sessions and both, left- and right-handed movements. In (a), (b) and (c) results are shown for the whole recording set of single-units. Subfigure (b) includes results for SUAs additionally checked for stability across trials. (a) Decoding power for different classification algorithms: PLDA, SVM with radial basis function kernel, SVM with linear kernel, multivariate Gaussian model, population vector approach (from left to right). Error bars depict the standard deviation. (b) Decoding power as a function of the width of the smoothing kernel. Colors and line styles as in Fig. 3a. (c) Accuracy of 2D trajectory prediction (average correlation coefficients across sessions and handedness) using either SVM regression (SVM) or the linear filter (LF). Supplementary Fig. 3. (jpg 15K)
Decoding power of individual single-units taken from the whole recording set regardless of stability. (a) Distribution of decoding power for individual SUA, shown for contralateral (left) and ipsilateral (right) movements. Dotted lines depict the chance level (0.125). (b) Contra- and ipsilateral decoding power was largely uncorrelated. Supplementary Methods (pdf 31K)
|
