Taku Yoshioka

Hierarchical Bayesian estimation for MEG inverse problem.

Source current estimation from MEG measurement is an ill-posed problem that requires prior assumptions about brain activity and an efficient estimation algorithm. In this article, we propose a new hierarchical Bayesian method introducing a hierarchical prior that can effectively incorporate both structural and functional MRI data. In our method, the variance of the source current at each source location is considered an unknown parameter and estimated from the observed MEG data and prior information by using the Variational Bayesian method. The fMRI information can be imposed as prior information on the variance distribution rather than the variance itself so that it gives a soft constraint on the variance. A spatial smoothness constraint, that the neural activity within a few millimeter radius tends to be similar due to the neural connections, can also be implemented as a hierarchical prior. The proposed method provides a unified theory to deal with the following three situations: (1) MEG with no other data, (2) MEG with structural MRI data on cortical surfaces, and (3) MEG with both structural MRI and fMRI data. We investigated the performance of our method and conventional linear inverse methods under these three conditions. Simulation results indicate that our method has better accuracy and spatial resolution than the conventional linear inverse methods under all three conditions. It is also shown that accuracy of our method improves as MRI and fMRI information becomes available. Simulation results demonstrate that our method appropriately resolves the inverse problem even if fMRI data convey inaccurate information, while the Wiener filter method is seriously deteriorated by inaccurate fMRI information.

Fast Gaussian process regression using representative data

Gaussian process regression is a Bayesian nonparametric regression model. Although the Gaussian process regression has shown good performance in various experiments, it suffers from O(N/sup 3/) computational cost, where N is the number of training data. We propose a method using representative data for the Gaussian process regression. The representative data are modified so that the regression model fits the original training data. The proposed method requires O(NM/sup 2/) computational cost, where M(<N) is the number of the representative data. According to our experiments, the results of the proposed method are comparable to those of the original method, although it requires only much smaller number of the representative data than the number of the original training data.

Clustering for time-series gene expression data using mixture of constrained PCAS

In a cluster analysis of gene expression time-series data, it is often required that genes with similar expression patterns should be classified into the same cluster regardless of their magnitude (scale). We propose a clustering method for gene expression time-series data based on mixture of constrained PCAs (MCPCA). The proposed method is scale-insensitive, while keeping the robustness to noise possibly involved in expression patterns with a small magnitude. We also propose a method that combines clustering results in order to improve the stability of the cluster analysis. The proposed method was applied to a time-series gene expression data set. In the experiment, an appropriate number of clusters was determined based on a statistical criterion. Furthermore, by combining clustering results, robustness of the cluster analysis was achieved. As a result, our method was able to catch biologically-meaningful expression patterns.

Clustering of Gene Expression Data by Mixture of PCA Models

Clustering techniques, such as hierarchical clustering, k- means algorithm and self-organizing maps, are widely used to analyze gene expression data. Results of these algorithms depend on several parameters, e.g., the number of clusters. However, there is no theoretical criterion to determine such parameters. In order to overcome this problem, we propose a method using mixture of PCA models trained by a variational Bayes (VB) estimation. In our method, good clustering results are selected based on the free energy obtained within the VB estimation. Furthermore, by taking an ensemble of estimation results, a robust clustering is achieved without any biological knowledge. Our method is applied to a clustering problem for gene expression data during a sporulation of Bacillus subtilis and it is able to capture characteristics of the sigma cascade.

A neural visualization method for WWW document clusters

Search engines are widely used for retrieving documents on the WWW. Visualization is useful for users to understand the retrieval results. When the retrieved documents are represented as document vectors, neural networks can be employed to visualize them. In this study, we consider the following two requirements for the visualization algorithm. One is that the cluster structure of document vectors is preserved. The other is that the visualization algorithm is fast. For these requirements, we employ basis function networks. Basis functions detect the cluster structure and weight parameters are adjusted by a fast algorithm so that the distance structure of the document vectors is preserved. Experiments show that our method is fast enough as an interface system.

Gaussian mixture prior distribution on artifactual current for MEG inverse problem

we observed individual neurons of the mouse brain using hard x-ray Talbottype phase-contrast micro-tomography with 1 m resolution at SPring-8. Furthermore, a nano-resolution hard x-ray Zernike-type phase-contrast microscope revealed nerve fibers and organelles including mitochondria and synapses in the neural tissue. In the near future, we will utilize that information to begin deciphering the wiring diagram of the brain by using the nano-resolution x-ray tomography.

Hierarchical Bayesian Inference of Brain Activity

Magnetoencephalography (MEG) can measure brain activity with millisecond-order temporal resolution, but its spatial resolution is poor, due to the ill-posed nature of the inverse problem, for estimating source currents from the electromagnetic measurement. Therefore, prior information on the source currents is essential to solve the inverse problem. We have proposed a new hierarchical Bayesian method to combine several sources of information. In our method, the variance of the source current at each source location is considered an unknown parameter and estimated from the observed MEG data and prior information by using variational Bayes method. The fMRI information can be imposed as prior distribution rather than the variance itself so that it gives a soft constraint on the variance. It is shown that the hierarchical Bayesian method has better accuracy and spatial resolution than conventional linear inverse methods by evaluating the resolution curve. The proposed method also demonstrated good spatial and temporal resolution for estimating current activity in early visual area evoked by a stimulus in a quadrant of the visual field.

Neural processing of audio-visual integration in speech perception: An MEG study

we observed individual neurons of the mouse brain using hard x-ray Talbottype phase-contrast micro-tomography with 1 m resolution at SPring-8. Furthermore, a nano-resolution hard x-ray Zernike-type phase-contrast microscope revealed nerve fibers and organelles including mitochondria and synapses in the neural tissue. In the near future, we will utilize that information to begin deciphering the wiring diagram of the brain by using the nano-resolution x-ray tomography.

22. Neural mechanisms associated with audio-visual integration in native Japanese and native English; an fMRI study

21. Time course of frontal lobe activation in schizophrenia: A multi-task study using two channel near-infrared spectroscopy— Masao Iwase, Michiyo Azechi, Koji Ikezawa, Ryouhei Ishii, Hidetoshi Takahashi, Takayuki Nakahachi, Leonides Canuet, Ryu Kurimoto, Hiroaki Kazui, Motoyuki Fukumoto, Naomi Iike, Kazutaka Ohi, Yuka Yasuda, Ryota Hashimoto, Masatoshi Takeda (Department of Psychiatry, Osaka University, Suita, Japan)

An analysis of MEG artifacts caused by heartbeat

Magnetoencephalography (MEG) directly measures the magnetic field caused by neural current activity, with a high temporal resolution. However, its amplitude is very weak and contaminated by various artifacts. One of the such an artifact is caused by heartbeat. In this study, we measured MEG and electrocardiogram (ECG) simultaneously. MEG was averaged with respect to an onset of ECG, that is, the peak of R-wave. Then, we applied equivalent current dipole (ECD) method to estimate current sources of artifacts caused by heartbeat. In addition, we propose a probabilistic model to remove such artifacts and applied to artificial data in order to confirm the efficiency of the method.