Unsupervised classification of neural spikes with a hybrid multilayer artificial neural network

The understanding of the brain structure and function and its computational style is one of the biggest challenges both in Neuroscience and Neural Computation. In order to reach this and to test the predictions of neural network modeling, it is necessary to observe the activity of neural populations. In this paper we propose a hybrid modular computational system for the spike classification of multiunits recordings. It works with no knowledge about the waveform, and it consists of two moduli: a Preprocessing (Segmentation) module, which performs the detection and centering of spike vectors using programmed computation; and a Processing (Classification) module, which implements the general approach of neural classification: feature extraction, clustering and discrimination, by means of a hybrid unsupervised multilayer artificial neural network (HUMANN). The operations of this artificial neural network on the spike vectors are: (i) compression with a Sanger Layer from 70 points vector to five principal component vector; (ii) their waveform is analyzed by a Kohonen layer; (iii) the electrical noise and overlapping spikes are rejected by a previously unreported artificial neural network named Tolerance layer; and (iv) finally the spikes are labeled into spike classes by a Labeling layer. Each layer of the system has a specific unsupervised learning rule that progressively modifies itself until the performance of the layer has been automatically optimized. The procedure showed a high sensitivity and specificity also when working with signals containing four spike types.     

Directional gene-transfer into the brain by an adenoviral vector tagged with magnetic nanoparticles

Adenoviral (Ad) vectors are useful for in vivo gene transfer into the brain. If Ad vectors are injected into the ventricle of mouse embryonic brain, Ad vectors introduce a foreign gene into neural progenitor cells on the surface of ventricle. However, Ad vectors were unable to deliver a foreign gene to a targeted region of the embryonic brain because Ad vectors evenly infected the neural progenitor cells on the surface of ventricle. Therefore, the Ad infection to the neural progenitor cells was uncontrollable. To develop a directional gene-transfer with Ad vector, we generated Ad vector tagged with magnetic nanoparticles (Ad-mag) by linking a biotinylated adenovirus vector with a streptavidin-conjugated magnetic nanoparticle. Ad-mags were attracted by magnetic force in vitro and in vivo. When Ad-mags were injected into the ventricle of mouse embryo and a strong magnet was attached to the head of the embryo, Ad-mags were attracted to the restricted direction or region where the magnet was placed. As a result, Ad-mags efficiently introduced a foreign gene into the restricted region of the brain.     

Immunology of neurological gene therapy: How T cells modulate viral vector-mediated therapeutic transgene expression through immunological synapses

Gene therapy has been shown to be a powerful new approach to the treatment of brain diseases. Brain neurodegenerations, brain tumors, inherited brain diseases, and autoimmune disorders are currently recognized as proper targets for gene therapeutics. Advances in the development of viral vectors (especially improvements in their immune profiles), the capacity to regulate transgene expression, and identification of appropriate therapeutic constructs have made the transition into clinical trials for gene therapy possible. One particular remaining challenge is the immune response that could be raised against either the viral vectors themselves or any regulatory or therapeutic transgenes. Because of the structure of brain immune responses, viral gene transfer into the brain can, under certain circumstances, be invisible to the systemic immune response and thus not generate a deleterious immune attack. If, however, the systemic immune system is primed against any vector antigen, the systemic immune response eliminates transgene expression and thus curtails the therapeutic efficacy of gene therapy. Mechanistic studies of brain immune responses indicate that the adaptive arm of the immune system may indeed be able to kill transduced cells. To move neurological gene therapy into the clinic in an effective and safe manner, these are the developments needed: novel viral vectors that either display a reduced capacity to stimulate an adaptive immune response or become invisible to the immune system after the delivery of the vector genome to the nucleus of transduced cells, and ways either to steer the immune response away from cytotoxic responses or to induce tolerance to gene therapy products.     

Transductive SVM for reducing the training effort in BCI

A brain-computer interface (BCI) provides a communication channel that translates human intention reflected by a brain signal such as electroencephalogram (EEG) into a control signal for an output device. In this work, the main concern is to reduce the training effort for BCI, which is often tedious and time consuming. Here we introduce a transductive support vector machines (TSVM) algorithm for the classification of EEG signals associated with mental tasks. TSVM possess the property of using both labeled and unlabeled data for reducing the calibration time in BCI and achieving good performance in classification accuracy. The advantages of the proposed method over the traditional supervised support vector machines (SVM) method are confirmed by about 2%-9% higher classification accuracies on a set of EEG recordings of three subjects from three-tasks-based mental imagery experiments.     

Single-trial classification of vowel speech imagery using common spatial patterns

With the goal of providing a speech prosthesis for individuals with severe communication impairments, we propose a control scheme for brain–computer interfaces using vowel speech imagery. Electroencephalography was recorded in three healthy subjects for three tasks, imaginary speech of the English vowels /a/ and /u/, and a no action state as control. Trial averages revealed readiness potentials at 200 ms after stimulus and speech related potentials peaking after 350 ms. Spatial filters optimized for task discrimination were designed using the common spatial patterns method, and the resultant feature vectors were classified using a nonlinear support vector machine. Overall classification accuracies ranged from 68% to 78%. Results indicate significant potential for the use of vowel speech imagery as a speech prosthesis controller.     

A preproenkephalin-neurofilament chimeric promoter in a helper virus-free herpes simplex virus vector enhances long-term expression in the rat striatum

Helper virus-free herpes simplex virus (HSV-1) plasmid vectors are an attractive system for gene transfer into neurons in the brain, but promoters that support long-term, neuronal-specific expression are required. Elucidation of general principles that govern long-term expression would likely assist efforts to develop improved promoters. Although expression from many promoters in HSV-1 vectors is unstable, two neuronal subtype-specific promoters, the preproenkephalin (ENK) promoter and the tyrosine hydroxylase (TH) promoter, support long-term expression. We have previously shown that 5' upstream sequences in the TH promoter are required for long-term expression, and addition of these upstream sequences to a neurofilament heavy gene (NF-H) promoter enhances long-term, neuronal-specific expression. The goal of this study was to determine if the upstream sequences from the TH promoter contain a unique element that enhances expression, or if other neuronal promoters also contain sequences that can enhance expression. To this end, we tested 5' upstream sequences in the ENK promoter. We isolated a vector that fuses upstream sequences from the ENK promoter to the NF-H promoter. This vector supported expression in the striatum for 2 months after gene transfer, the longest time point evaluated. Expression was neuronal specific. As ENK and TH are a peptide neurotransmitter and a classical neurotransmitter biosynthetic enzyme, respectively, these results suggest that a significant number of promoters for neurotransmitter biosynthetic genes may contain elements that can enhance expression from HSV-1 vectors. The strategy of using upstream sequences from neuronal subtype-specific promoters to enhance expression from heterologous promoters is discussed.     

A divide-and-combine method for large scale nonparallel support vector machines

Nonparallel Support Vector Machine (NPSVM) which is more flexible and has better generalization than typical SVM is widely used for classification. Although some methods and toolboxes like SMO and libsvm for NPSVM are used, NPSVM is hard to scale up when facing millions of samples. In this paper, we propose a divide-and-combine method for large scale nonparallel support vector machine (DCNPSVM). In the division step, DCNPSVM divide samples into smaller sub-samples aiming at solving smaller subproblems independently. We theoretically and experimentally prove that the objective function value, solutions, and support vectors solved by DCNPSVM are close to the objective function value, solutions, and support vectors of the whole NPSVM problem. In the combination step, the sub-solutions combined as initial iteration points are used to solve the whole problem by global coordinate descent which converges quickly. In order to balance the accuracy and efficiency, we adopt a multi-level structure which outperforms state-of-the-art methods. Moreover, our DCNPSVM can tackle unbalance problems efficiently by tuning the parameters. Experimental results on lots of large data sets show the effectiveness of our method in memory usage, classification accuracy and time consuming.     

Enhanced active segment selection for single-trial EEG classification

In this study, an electroencephalogram (EEG) analysis system is proposed for single-trial classification of both motor imagery (MI) and finger-lifting EEG data. Applying event-related brain potential (ERP) data acquired from the sensorimotor cortices, the system mainly consists of three procedures; enhanced active segment selection, feature extraction, and classification. In addition to the original use of continuous wavelet transform (CWT) and Student 2-sample t statistics, the two-dimensional (2D) anisotropic Gaussian filter further refines the selection of active segments. The multiresolution fractal features are then extracted from wavelet data by using proposed modified fractal dimension. Finally, the support vector machine (SVM) is used for classification. Compared to original active segment selection, with several popular features and classifier on both the MI and finger-lifting data from 2 data sets, the results indicate that the proposed method is promising in EEG classification.     

Sendai virus vector-mediated brain-derived neurotrophic factor expression ameliorates memory deficits and synaptic degeneration in a transgenic mouse model of Alzheimer's disease

Growing evidence suggests that decreased brain-derived neurotrophic factor (BDNF) levels are associated with Alzheimer's disease (AD) pathogenesis. Therefore, BDNF gene therapy is considered to be a promising therapeutic strategy for treating AD. Sendai virus (SeV) is a type I parainfluenza virus that does not interact with host chromosomes because of its strict cytoplasmic life cycle. Although SeV is nonpathogenic in primates, including humans, its infectivity for neurons is strong. Here we demonstrate that SeV vectors effectively infected neurons, even though they were injected into subcortical white matter. Moreover, SeV vectors significantly induced BDNF expression, ameliorating synaptic degeneration and memory deficits in a transgenic mouse model of AD (Tg2576). This is the first study to demonstrate that viral vector administration in white matter is sufficient to restore cognitive function in vivo. These results also support the feasibility of using SeV vectors for gene therapy targeting the brain.     

A system, using neural cell lines, to characterize HSV-1 vectors containing genes which affect neuronal physiology, or neuronal promoters.

Among the potential uses of defective herpes simplex virus (HSV-1) vectors are to study neuronal physiology, neuronal gene regulation, and to perform gene therapy of neuronal diseases. The prototype HSV-1 vector, pHSVlac, stably expresses Escherichia coli beta-galactosidase from the HSV-1 immediate early (IE) 4/5 promoter in cultured rat peripheral and CNS neurons, and in neurons in the adult rat brain. The LacZ gene and the IE 4/5 promoter in pHSVlac can be replaced with genes which affect neuronal physiology or cellular promoters, respectively. A system is required to characterize these HSV-1 vectors; cultured neurons, a mixture of different kinds of neurons and glia, cannot be used. In contrast, neural cell lines represent a homogenous population of neural cells available in virtually unlimited quantities. A system, using neural cell lines, to characterize HSV-1 vectors carrying other genes or promoters is now reported: First, 4 assays are described to detect HSV-1 vector DNA, RNA transcribed from the vector, and to quantitate beta-galactosidase expression. Second, 8 cell lines derived from rodents, primates, and humans were infected with pHSVlac virus and shown to express beta-galactosidase. The cell lines tested included adrenergic and cholinergic mouse neuroblastoma cells, rat pheochromocytoma cells, rodent pituicytes, and human neuroblastoma cells. Infection of these cell lines should prove useful for characterizing HSV-1 vectors with molecular and biochemical assays. Third, differentiated rat pheochromocytoma and mouse neuroblastoma cells, which resemble neurons, were infected with pHSVlac virus and shown to stably express beta-galactosidase. Infection of these cells should be useful for determining the effect of various HSV-1 vectors on neuronal physiology. Thus, HSV-1 vectors containing various genes or promoters can be characterized using the system described in this study.     

Regulated protein expression for in vivo gene therapy for neurological disorders: progress, strategies, and issues

The field of in vivo gene therapy has matured to the point where there are numerous clinical trials underway including late-stage clinical trials. Several viral vectors are especially efficient and support lifetime protein expression in the brain and a number of clinical trials are underway for various progressive or chronic neurological disorders including Parkinson's disease, Alzheimer's disease, and Batten's disease. To date, however, none of the vectors in clinical use have any direct way to reverse or control their transgene product in the event continued protein expression should become problematic. Several schemes that use elements within the vector design have been developed that allow an external drug or pro-drug to alter ongoing protein expression after in vivo gene transfer. The most promising and most studied regulated protein expression methods for in vivo gene transfer are reviewed. In addition, potential scientific and clinical advantages of transgene regulation for gene therapy are discussed.     

The use of a recombinant lentiviral vector for ex vivo gene transfer into the rat CNS

A major obstacle in ex vivo gene transfer has been the loss of transgene expression soon after implantation of the grafted transduced cells. Recently, a lentiviral vector system has been developed which has proven to express high levels of transgenes in vivo after direct injection into the tissue. In this study, we have investigated the use of such a vector for ex vivo gene transfer to the brain. A number of neural cell types were found to be permissive to transduction by the lentiviral vector in vitro and a majority of them expressed the transgene after transplantation to the rat brain. Transgene expression was detected up to 8 weeks post-grafting. These findings suggest that recombinant lentiviral vectors may be used for further development of ex vivo gene therapy protocols to the CNS.     

Prevention of neuropathology in the mouse model of Hurler syndrome

A defect of the lysosomal enzyme alpha-L-iduronidase (IDUA) interrupts heparan and dermatan sulfate degradation and causes neuropathology in children with severe forms of mucopolysaccharidosis type I (MPSI, Hurler syndrome). Enzyme substitution therapy is beneficial but ineffective on the central nervous system. We could deliver the missing enzyme to virtually the entire brain of MPSI mice through a single injection of gene transfer vectors derived from adenoassociated virus serotype 2 (AAV2) or 5 (AAV5) coding for human IDUA. This result was reproducibly achieved with both vector types in 46 mice and persisted for at least 26 weeks. Success was more frequent, enzyme activity was higher, and corrected areas were broader with AAV5 than with AAV2 vectors. Treatment presumably reversed and certainly prevented the accumulation of GM2 and GM3 gangliosides, which presumably participates to neuropathology. Lysosomal distension, which already was present at the time of treatment, had disappeared from both brain hemispheres and was minimal in the cerebellum in mice analyzed 26 weeks after injection. This study shows that pathology associated with MPSI can be prevented in the entire mouse brain by a single AAV vector injection, providing a preliminary evaluation of the feasibility of gene therapy to stop neuropathology in Hurler syndrome.     

Epileptic EEG classification based on extreme learning machine and nonlinear features

The automatic detection and classification of epileptic EEG are significant in the evaluation of patients with epilepsy. This paper presents a new EEG classification approach based on the extreme learning machine (ELM) and nonlinear dynamical features. The theory of nonlinear dynamics has been a powerful tool for understanding brain electrical activities. Nonlinear features extracted from EEG signals such as approximate entropy (ApEn), Hurst exponent and scaling exponent obtained with detrended fluctuation analysis (DFA) are employed to characterize interictal and ictal EEGs. The statistics indicate that the differences of those nonlinear features between interictal and ictal EEGs are statistically significant. The ELM algorithm is employed to train a single hidden layer feedforward neural network (SLFN) with EEG nonlinear features. The experiments demonstrate that compared with the backpropagation (BP) algorithm and support vector machine (SVM), the performance of the ELM is better in terms of training time and classification accuracy which achieves a satisfying recognition accuracy of 96.5% for interictal and ictal EEG signals.     

Targeted gene transfer to nigrostriatal neurons in the rat brain by helper virus-free HSV-1 vector particles that contain either a chimeric HSV-1 glycoprotein C-GDNF or a gC-BDNF protein

Direct gene transfer into neurons has potential for both studying neuronal physiology and for developing gene therapy treatments for specific neurological conditions. Due to the heterogeneous cellular composition of the brain, cell-type-specific recombinant gene expression is required for many potential applications of neuronal gene transfer. The two prevalent approaches for achieving cell-type-specific expression are to use a cell-type-specific promoter to control recombinant gene expression or to modify a virus vector particle to target gene transfer to a specific cell type. Targeted gene transfer to multiple peripheral cell types has been described, but targeted gene transfer to a specific type of neuron in the brain has yet to be reported. Targeted gene transfer approaches with Herpes Simplex Virus (HSV-1) vectors have focused on modifying glycoprotein C (gC) to remove the heparin binding domain and add a binding activity for a specific protein on the cell surface. This study was designed to develop HSV-1 vectors that target gene transfer to cells that contain receptors for either glial-cell-line-derived neurotrophic factor (GDNF) or brain-derived neurotrophic factor (BDNF), such as nigrostriatal neurons. We isolated chimeric gC-GDNF or chimeric gC-BDNF constructs, and the resulting proteins were incorporated into HSV-1 virus particles. We performed helper virus-free HSV-1 vector packaging in the presence of each chimeric protein. The resulting vector stocks supported 2.2- to 5.0-fold targeted gene transfer to nigrostriatal neurons in the rat brain, compared to vector particles that contained wild-type (wt) gC. Gene transfer to nigrostriatal neurons by vector particles that contained chimeric gC-BDNF was reduced by preincubation with an anti-BDNF antibody. Targeted gene transfer to neurons that contain specific neurotrophic factor receptors may benefit specific physiological and gene therapy studies.     

Lentiviral vectors express chondroitinase ABC in cortical projections and promote sprouting of injured corticospinal axons

Several diseases and injuries of the central nervous system could potentially be treated by delivery of an enzyme, which might most effectively be achieved by gene therapy. In particular, the bacterial enzyme chondroitinase ABC is beneficial in animal models of spinal cord injury. We have adapted the chondroitinase gene so that it can direct secretion of active chondroitinase from mammalian cells, and inserted it into lentiviral vectors. When injected into adult rat brain, these vectors lead to extensive secretion of chondroitinase, both locally and from long-distance axon projections, with activity persisting for more than 4 weeks. In animals which received a simultaneous lesion of the corticospinal tract, the vector reduced axonal die-back and promoted sprouting and short-range regeneration of corticospinal axons. The same beneficial effects on damaged corticospinal axons were observed in animals which received the chondroitinase lentiviral vector directly into the vicinity of a spinal cord lesion.     

Helper virus-free HSV-1 vectors packaged both in the presence of VSV G protein and in the absence of HSV-1 glycoprotein B support gene transfer into neurons in the rat striatum

Herpes simplex virus (HSV-1) vectors have potential for gene transfer into quiescent cells, but the gene transfer process could be more efficient. In other vector systems, both the titers and the efficiency of gene transfer have been enhanced by pseudotyping the vector particles with vesicular stomatitis virus (VSV) G protein. In this report, we pseudotyped helper virus-free HSV-1 plasmid vectors with VSV G protein. Packaging was performed in the presence of both VSV G protein and a deletion in an essential HSV-1 glycoprotein, gB. The resulting vector stocks supported gene transfer into both fibroblast and neuronal cell lines. VSV G protein was required for gene transfer because preincubation of these vector stocks with antibodies directed against either VSV G protein or VSV reduced the titer to undetectable levels. Although the titers were lower than those obtained using the unmodified vector system, the titers were not increased by use of chimeric proteins that contain the extracellular domain of VSV G protein and the transmembrane and/or cytoplasmic domains of specific HSV-1 glycoproteins. Also, the titers were not increased by performing the packaging in the presence of deletions in multiple HSV-1 glycoproteins. Nonetheless, pHSVlac pseudotyped with VSV G protein supported gene transfer into striatal neurons in the rat brain. Thus, HSV-1 vectors pseudotyped with VSV G protein may be useful for specific gene transfer studies.     

Amplitude and phase coupling measures for feature extraction in an EEG-based brain-computer interface

Most of the feature extraction methods in existing brain-computer interfaces (BCIs) are based on the dynamic behavior of separate signals, without using the coupling information between different brain regions. In this paper, amplitude and phase coupling measures, quantified by a nonlinear regressive coefficient and phase locking value respectively, were used for feature extraction. The two measures were based on three different coupling methods determined by neurophysiological a priori knowledge, and applied to a small number of electrodes of interest, leading to six feature vectors for classification. Five subjects participated in an online BCI experiment during which they were asked to imagine a movement of either the left or right hand. The electroencephalographic (EEG) recordings from all subjects were analyzed offline. The averaged classification accuracies of the five subjects ranged from 87.4% to 92.9% for the six feature vectors and the best classification accuracies of the six feature vectors ranged between 84.4% and 99.6% for the five subjects. The performance of coupling features was compared with that of the autoregressive (AR) feature. Results indicated that coupling measures are appropriate methods for feature extraction in BCIs. Furthermore, the combination of coupling and AR feature can effectively improve the classification accuracy due to their complementarities.