Visualization Portfolio of Blue Sheffer


Brain-Machine Interface (BMI) Cursor Perturbation Experiment

This visualization uses data from the Neural Prosthetic Systems Laboratory (Shenoy Lab) at Stanford. The work is an extension of a 2015 NER conference paper by Stavisky et al. in which a step cursor position offset was applied while a monkey controlled a BMI cursor using decoded motor cortical spiking activity. The task is to decode the direction of the offset (either left of right) by looking at solely the neural activity.

Each trial of the experiment can be viewed as a point in high dimensional neural firing rate "state space" where each dimension corresponds to a single unit of a microelectrode array. Here, we project the trials into 3D space using principal component analysis, where the components were generated from the trial averaged difference between activity on left trials and right trials. Because the left and right trials are clearly separable in this 3D space, even simple classification algorithms such as support vector machines and logistic regression can classifiy the direction of the offset with high accuracy. The cursor trajectory for each trial is plotted on the right, with the black dot as the cursor at the given time. Click on the plots to view individual trials evolve in the state space. Allow ~10 seconds for the vizualization to load.

This visualization would not have been possible without Dave Deriso (CME 161 instructor), Juhi Naik (CME 161 TA), and the wonderful team at the Shenoy Lab (esp. Sergey Stavisky).

Time (ms):