Shiro Usui

Digital Low-Pass Differentiation for Biological Signal Processing

Digital low-pass differentiation is often required in processing various biological or biomechanical data. However, both the nature of biological signals and the use of micro-or minicomputers in such applications imply the need for simple, low-order, and fast differentiation methods, rather than sophisticated high-order algorithms. Responding to this need, we investigate here the low-pass first- and second-order digital differentiation from both theoretical and practical points of view, in order to achieve good and simple algorithms. In contrast with most of the research works previously done in this field, whose main aim was to achieve better accuracy even in the cost of using quite high-order algorithms, we restrict ourselves in this study only to low orders, being interested not only in the accuracy achieved, but also in the simplicity of the algorithm. After discussing the theoretical considerations concerning our optimum low-pass differentiation filters, we present our simple low-order filters and show them to be not only very convenient for use, but also almost optimum.

Identification of Retinal Ganglion Cells and Their Projections Involved in Central Transmission of Information about Upward and Downward Image Motion

The direction of image motion is coded by direction-selective (DS) ganglion cells in the retina. Particularly, the ON DS ganglion cells project their axons specifically to terminal nuclei of the accessory optic system (AOS) responsible for optokinetic reflex (OKR). We recently generated a knock-in mouse in which SPIG1 (SPARC-related protein containing immunoglobulin domains 1)-expressing cells are visualized with GFP, and found that retinal ganglion cells projecting to the medial terminal nucleus (MTN), the principal nucleus of the AOS, are comprised of SPIG1+ and SPIG1− ganglion cells distributed in distinct mosaic patterns in the retina. Here we examined light responses of these two subtypes of MTN-projecting cells by targeted electrophysiological recordings. SPIG1+ and SPIG1− ganglion cells respond preferentially to upward motion and downward motion, respectively, in the visual field. The direction selectivity of SPIG1+ ganglion cells develops normally in dark-reared mice. The MTN neurons are activated by optokinetic stimuli only of the vertical motion as shown by Fos expression analysis. Combination of genetic labeling and conventional retrograde labeling revealed that axons of SPIG1+ and SPIG1− ganglion cells project to the MTN via different pathways. The axon terminals of the two subtypes are organized into discrete clusters in the MTN. These results suggest that information about upward and downward image motion transmitted by distinct ON DS cells is separately processed in the MTN, if not independently. Our findings provide insights into the neural mechanisms of OKR, how information about the direction of image motion is deciphered by the AOS.

NEUROINFORMATICS: THE INTEGRATION OF SHARED DATABASES AND TOOLS TOWARDS INTEGRATIVE NEUROSCIENCE

There is significant interest amongst neuroscientists in sharing neuroscience data and analytical tools. The exchange of neuroscience data and tools between groups affords the opportunity to differently re-analyze previously collected data, encourage new neuroscience interpretations and foster otherwise uninitiated collaborations, and provide a framework for the further development of theoretically based models of brain function. Data sharing will ultimately reduce experimental and analytical error. Many small Internet accessible database initiatives have been developed and specialized analytical software and modeling tools are distributed within different fields of neuroscience. However, in addition large-scale international collaborations are required which involve new mechanisms of coordination and funding. Provided sufficient government support is given to such international initiatives, sharing of neuroscience data and tools can play a pivotal role in human brain research and lead to innovations in neuroscience, informatics and treatment of brain disorders. These innovations will enable application of theoretical modeling techniques to enhance our understanding of the integrative aspects of neuroscience. This article, authored by a multinational working group on neuroinformatics established by the Organization for Economic Co-operation and Development (OECD), articulates some of the challenges and lessons learned to date in efforts to achieve international collaborative neuroscience.

Neuroscience data and tool sharing: a legal and policy framework for neuroinformatics.

The requirements for neuroinformatics to make a significant impact on neuroscience are not simply technical—the hardware, software, and protocols for collaborative research—they also include the legal and policy frameworks within which projects operate. This is not least because the creation of large collaborative scientific databases amplifies the complicated interactions between proprietary, for-profit R&D and public “open science.” In this paper, we draw on experiences from the field of genomics to examine some of the likely consequences of these interactions in neuroscience.Facilitating the widespread sharing of data and tools for neuroscientific research will accelerate the development of neuroinformatics. We propose approaches to overcome the cultural and legal barriers that have slowed these developments to date. We also draw on legal strategies employed by the Free Software community, in suggesting frame-works neuroinformatics might adopt to reinforce the role of public-science databases, and propose a mechanism for identifying and allowing “open science” uses for data whilst still permitting flexible licensing for secondary commercial research.

Reconstruction of Munsell color space by a five-layer neural network

We have constructed a wine-glass-type five-layer neural network and generated an identity mapping of the surface spectral-reflectance data of 1280 Munsell color chips, using a backpropagation learning algorithm. To achieve an identity mapping, the same data set is used for the input and for the teacher. After the learning was completed, we analyzed the responses to individual chips of the three hidden units in the middle layer in order to obtain the internal representation of the color information. We found that each of the three hidden units corresponds to a psychological color attribute, that is, the Munsell value (luminance), red–green, and yellow–blue. We also examined the relationship between the internal representation and the number of hidden units and found that the network with three hidden units acquires optimum color representation. The five-layer neural network is shown to be an efficient method for reproducing the transformation of color information (or color coding) in the visual system.

Kick-out learning algorithm to reduce the oscillation of weights

Abstract The back-propagation algorithm, when used with an unmodified gradient descent term, converges very slowly, because the weights oscillate in regions where the error surface forms a ravine. To improve the convergence, the momentum term was introduced. However, the effect of that term on the reduction of oscillations has been insufficiently considered. In this paper, we point out that this term has not been effective in reducing the oscillation. To overcome the oscillations, we focus on the very bottom of a ravine where the direction of steepest descent is the same as the downward direction along the ravine bottom. We describe a method to correct the value of the weights near the bottom of a ravine and propose a new acceleration algorithm based on that correction. The distinctive feature is the correction term that uses the difference of gradients that is invoked during the oscillation. We show that, using the proposed algorithm, the convergence speed is substantially improved in ravine regions.

A model for nonlinear stochastic behavior of the pupil

In this paper we present new experimental results that show pupillary noise to be multiplicative in a particular fashion with greatest variance in midrange and smaller variance at high and low ranges. This confirms the finding of multiplicative noise by Stanten and Stark (1966), but modifies and extends the relationship they suggested between standard deviation and mean pupil diameter. We propose a parametric model of the iris muscle which not only describes the static characteristics of pupil response to given stimuli, but also explains its random fluctuations in terms of probability density functions. We emphasize the point that the range nonlinearity is not due to decreased gain at the extrema of the pupil range, but is operational over a wide portion of the pupillary behavioral range, hence its name — “expansive range nonlinearity”. We conclude that noise amplitude, which is a function of the pupil diameter, closely parallels the changes in deterministic gain. Thus pupil noise can be simply considered as cross-talk additive Gaussian noise injected into the pupil system at midbrain level.

On the problem of applying AIC to determine the structure of a layered feedforward neural network

AIC (Akaikes information criterion) has been thought to be effective to determine an optimal structure of layered feedforward neural networks. However, it has not been clarified from the theoretical point of view. On the other hand, it is known that a connection weight of the network can be nonunique in some cases. In this paper, we show that AIC can not be derived for three-layered networks due to the nonuniqueness of the connection weight. Through numerical simulations of data fitting with three-layered neural networks, we show that the structure determined by AIC tends to be more complex because of the inherent data fitting capability of the network.

Estimation of autonomic nervous activity using the inverse dynamic model of the pupil muscle plant.

In order to elucidate the mechanism of the pupillary control system, the internal property of the pupillary muscle plant, as well as the autonomic nervous input to the muscle plant, must be analyzed. In this study, we approach the problem first by constructing a new homeomorphic biomechanical model for the human pupillary muscle plant (forward dynamic model). We showed that the model is able not only to reproduce various experimental results that exhibit various nonlinearities but also to explain how such nonlinear responses are generated in terms of the internal property of the model. Then, we contrive a possible method to estimate the autonomic nervous input to the muscle plant. This method utilizes the inverse dynamic model of the pupillary muscle plant so that the autonomic nervous input can be estimated from the pupillary response. We applied this method to the experimental step responses, and showed that the estimated neural input indicates characteristics quite similar to the results of the physiological experiment. Last, we discuss the origin of the pupillary escape and capture as well as the sustained and transient components of the pupillary response, based on the analysis of the forward and/or inverse dynamic model.

Ionic current model of the vertebrate rod photoreceptor.

We describe voltage- and calcium-dependent ionic currents in the photoreceptor inner segments similar to the Hodgkin and Huxley (Journal of Physiology, 117, 500-544, 1952) equations. The model is used to describe both rods and cones by adjusting parameters. To simulate the light response, the inner segment model was connected with the phototransduction model proposed by Torre et al. (Cold Spring Harbor Symposia on Quantitative Biology, 55, 563-573, 1990). The role of individual ionic currents in the inner segment in shaping the light response was analyzed through computer simulations. The results suggest that: (1) the transient hyperpolarization to a bright flash is generated by Ih; (2) the oscillation after prolonged hyperpolarization in rods results from the interaction among Ica, IK(Ca), and ICI(Ca). Since the present model describes the biophysical processes from phototransduction to voltage response, the model can be used for analyzing the light response properties of the photoreceptors quantitatively.