%0 Journal Article
%T Ensemble Fractional Sensitivity: A Quantitative Approach to Neuron Selection for Decoding Motor Tasks
%A Girish Singhal
%A Vikram Aggarwal
%A Soumyadipta Acharya
%A Jose Aguayo
%A Jiping He
%A Nitish Thakor
%J Computational Intelligence and Neuroscience
%D 2010
%I Hindawi Publishing Corporation
%R 10.1155/2010/648202
%X A robust method to help identify the population of neurons used for decoding motor tasks is developed. We use sensitivity analysis to develop a new metric for quantifying the relative contribution of a neuron towards the decoded output, called ˇ°fractional sensitivity.ˇ± Previous model-based approaches for neuron ranking have been shown to largely depend on the collection of training data. We suggest the use of an ensemble of models that are trained on random subsets of trials to rank neurons. For this work, we tested a decoding algorithm on neuronal data recorded from two male rhesus monkeys while they performed a reach to grasp a bar at three orientations ( , , or ). An ensemble approach led to a statistically significant increase of 5% in decoding accuracy and 25% increase in identification accuracy of simulated noisy neurons, when compared to a single model. Furthermore, ranking neurons based on the ensemble fractional sensitivities resulted in decoding accuracies 10%¨C20% greater than when randomly selecting neurons or ranking based on firing rates alone. By systematically reducing the size of the input space, we determine the optimal number of neurons needed for decoding the motor output. This selection approach has practical benefits for other BMI applications where limited number of electrodes and training datasets are available, but high decoding accuracies are desirable. 1. Introduction A Brain-Machine Interface (BMI) uses activities recorded from various motor areas, such as the primary motor, premotor and posterior parietal cortex, to translate neural activities recorded from the brain into commands to control an external device. Traditionally, BMI researchers have used extracellular action potentials from localized cortical sites, primarily in the motor cortex, to provide closed-loop control of a computer cursor [1] or a robotic arm in 3D space [2¨C4]. More recently, researchers have now begun to use implanted microelectrode arrays, which can simultaneously sample neuronal ensembles from various cortical sites [5¨C7]. These electrode arrays are surgically placed in cortical regions which are correlated to the motor function. The relevant cortical regions are identified using anatomical guidance, preliminary probing of neural activity and imaging techniques such as FMRI. However, in multichannel recordings only 30%¨C40% of single units are typically relevant to the motor task [8]; the remaining neurons are either noisy or not task-related. This adds uncorrelated dimensions to the input space, thereby degrading the predictive performance of the
%U http://www.hindawi.com/journals/cin/2010/648202/