Estimation of hyper-parameters for a hierarchical model of combined cortical and extra-brain current sources in the MEG inverse problem.

One of the major obstacles in estimating cortical currents from MEG signals is the disturbance caused by magnetic artifacts derived from extra-cortical current sources such as heartbeats and eye movements. To remove the effect of such extra-brain sources, we improved the hybrid hierarchical variational Bayesian method (hyVBED) proposed by Fujiwara et al. (NeuroImage, 2009). hyVBED simultaneously estimates cortical and extra-brain source currents by placing dipoles on cortical surfaces as well as extra-brain sources. This method requires EOG data for an EOG forward model that describes the relationship between eye dipoles and electric potentials. In contrast, our improved approach requires no EOG and less a priori knowledge about the current variance of extra-brain sources. We propose a new method, "extra-dipole," that optimally selects hyper-parameter values regarding current variances of the cortical surface and extra-brain source dipoles. With the selected parameter values, the cortical and extra-brain dipole currents were accurately estimated from the simulated MEG data. The performance of this method was demonstrated to be better than conventional approaches, such as principal component analysis and independent component analysis, which use only statistical properties of MEG signals. Furthermore, we applied our proposed method to measured MEG data during covert pursuit of a smoothly moving target and confirmed its effectiveness.

Group analysis and classification of working memory task conditions using electroencephalogram cortical currents during an n-back task.

Electroencephalographic studies of working memory have demonstrated cortical activity and oscillatory representations without clarifying how the stored information is retained in the brain. To address this gap, we measured scalp electroencephalography data, while participants performed a modified n-back working memory task. We calculated the current intensities from the estimated cortical currents by introducing a statistical map generated using Neurosynth as prior information. Group analysis of the cortical current level revealed that the current amplitudes and power spectra were significantly different between the modified n-back and delayed match-to-sample conditions. Additionally, we classified information on the working memory task conditions using the amplitudes and power spectra of the currents during the encoding and retention periods. Our results indicate that the representation of executive control over memory retention may be mediated through both persistent neural activity and oscillatory representations in the beta and gamma bands over multiple cortical regions that contribute to visual working memory functions.

Common cortical areas have different neural mechanisms for covert and overt visual pursuits.

Although humans can direct their attention to visual targets with or without eye movements, it remains unclear how different brain mechanisms control visual attention and eye movements together and/or separately. Here, we measured MEG and fMRI data during covert/overt visual pursuit tasks and estimated cortical currents using our previously developed extra-dipole, hierarchical Bayesian method. Then, we predicted the time series of target positions and velocities from the estimated cortical currents of each task using a sparse machine-learning algorithm. The predicted target positions/velocities had high temporal correlations with actual visual target kinetics. Additionally, we investigated the generalization ability of predictive models among three conditions: control, covert, and overt pursuit tasks. When training and testing data were the same tasks, the largest reconstructed accuracies were overt, followed by covert and control, in that order. When training and testing data were selected from different tasks, accuracies were in reverse order. These results are well explained by the assumption that predictive models consist of combinations of three computational brain functions: visual information-processing, maintenance of attention, and eye-movement control. Our results indicate that separate subsets of neurons in the same cortical regions control visual attention and eye movements differently.

Practice reduces task relevant variance modulation and forms nominal trajectory.

Humans are capable of achieving complex tasks with redundant degrees of freedom. Much attention has been paid to task relevant variance modulation as an indication of online feedback control strategies to cope with motor variability. Meanwhile, it has been discussed that the brain learns internal models of environments to realize feedforward control with nominal trajectories. Here we examined trajectory variance in both spatial and temporal domains to elucidate the relative contribution of these control schemas. We asked subjects to learn reaching movements with multiple via-points, and found that hand trajectories converged to stereotyped trajectories with the reduction of task relevant variance modulation as learning proceeded. Furthermore, variance reduction was not always associated with task constraints but was highly correlated with the velocity profile. A model assuming noise both on the nominal trajectory and motor command was able to reproduce the observed variance modulation, supporting an expression of nominal trajectories in the brain. The learning-related decrease in task-relevant modulation revealed a reduction in the influence of optimal feedback around the task constraints. After practice, the major part of computation seems to be taken over by the feedforward controller around the nominal trajectory with feedback added only when it becomes necessary.

Feedforward impedance control efficiently reduce motor variability.

Despite the existence of neural noise, which leads variability in motor commands, the central nervous system can effectively reduce movement variance at the end effector to meet task requirements. Although online correction based on feedback information is essential for reducing error, feedforward impedance control is another way to regulate motor variability. This Update Article reviews key studies examining the relation between task constraints and impedance control for human arm movement. When a smaller reaching target is given as a task constraint, flexor and extensor muscles are co-activated, and positional variance is decreased around the task constraint. Trial-by-trial muscle activations revealed no on-line feedback correction, indicating that humans are able to regulate their impedance in advance. These results demonstrate that not only on-line feedback correction, but also feedforward impedance control, helps reduce the motor variability caused by internal noise to realize dexterous movements of human arms. A computational model of movement planning considering the presence of signal-dependent noise provides a unifying framework that potentially accounts for optimizing impedance to maximize accuracy. A recently proposed learning algorism formulated as a V-shaped learning function explains how the central nervous system acquires impedance to optimize accuracy as well as stability and efficiency.