Classification Of Neural Signals Research Articles

A HYBRID MODEL FOR THE CLASSIFICATION OF PHYSIOLOGICAL AND NEURAL SIGNAL USING CNN-LSTM TECHNIQUE

Physiological and neural signal classification has many important applications in healthcare, including medical diagnosis and monitoring. For example, electrocardiogram (ECG) classification can be used to detect arrhythmias and other cardiac abnormalities; while EEG classification can be used to diagnose neurological disorders such as epilepsy and sleep apnea. This paper presents an LSTM model for the decoding of physiological and neural signals. In this paper, an electroencephalography brain signal data which was gotten from Kaggle.com was used. The dataset was pre-processed so as to remove noise from the data. The pre-processed data was used in training the LSTM model. The LSTM model was trained on fourteen (14) steps. The result of the LSTM model showed an accuracy of 85% at the first step and a validation (testing) accuracy of 90%. For the fourteenth step, the model achieved an accuracy result of 98% for training and 94% for validation (testing). We also evaluated the performance of the model using a classification report and confusion matrix. The result of the classification report showed an accuracy of 95%, which is implication that the performance of the model on the test data is efficient. The confusion matrix was used to specify how well the proposed model classified the electroencephalography signal. The result of the confusion matrix showed that the model predicted the result correctly to be neutral 151 out of 153, positive to be 127 out of 142, and negative to be 128 out of 132. The result showed that the level of false positive and negative values is minimal (0.02% and 0.05%).

Decoding Speech from Single Trial MEG Signals Using Convolutional Neural Networks and Transfer Learning.

Decoding speech directly from the brain has the potential for the development of the next generation, more efficient brain computer interfaces (BCIs) to assist in the communication of patients with locked-in syndrome (fully paralyzed but aware). In this study, we have explored the spectral and temporal features of the magnetoencephalography (MEG) signals and trained those features with convolutional neural networks (CNN) for the classification of neural signals corresponding to phrases. Experimental results demonstrated the effectiveness of CNNs in decoding speech during perception, imagination, and production tasks. Furthermore, to overcome the long training time issue of CNNs, we leveraged principal component analysis (PCA) for spatial dimension reduction of MEG data and transfer learning for model initialization. Both PCA and transfer learning were found to be highly beneficial for faster model training. The best configuration (50 principal coefficients + transfer learning) led to more than 10 times faster training than the original setting while the speech decoding accuracy remained at a similarly high level.

Neural signals evoked by stimuli of increasing social scene complexity are detectable at the single-trial level and right lateralized.

Classification of neural signals at the single-trial level and the study of their relevance in affective and cognitive neuroscience are still in their infancy. Here we investigated the neurophysiological correlates of conditions of increasing social scene complexity using 3D human models as targets of attention, which may also be important in autism research. Challenging single-trial statistical classification of EEG neural signals was attempted for detection of oddball stimuli with increasing social scene complexity. Stimuli had an oddball structure and were as follows: 1) flashed schematic eyes, 2) simple 3D faces flashed between averted and non-averted gaze (only eye position changing), 3) simple 3D faces flashed between averted and non-averted gaze (head and eye position changing), 4) animated avatar alternated its gaze direction to the left and to the right (head and eye position), 5) environment with 4 animated avatars all of which change gaze and one of which is the target of attention. We found a late (> 300 ms) neurophysiological oddball correlate for all conditions irrespective of their complexity as assessed by repeated measures ANOVA. We attempted single-trial detection of this signal with automatic classifiers and obtained a significant balanced accuracy classification of around 79%, which is noteworthy given the amount of scene complexity. Lateralization analysis showed a specific right lateralization only for more complex realistic social scenes. In sum, complex ecological animations with social content elicit neurophysiological events which can be characterized even at the single-trial level. These signals are right lateralized. These finding paves the way for neuroscientific studies in affective neuroscience based on complex social scenes, and given the detectability at the single trial level this suggests the feasibility of brain computer interfaces that can be applied to social cognition disorders such as autism.

Classification of neural signals from sparse autoregressive features

This paper introduces a signal classification framework that can be used for brain–computer interface design. The actual classification is performed on sparse autoregressive features. It can use any well-known classification algorithm, such as discriminant analysis, linear logistic regression and support vector machines. The autoregressive coefficients of all signals and channels are simultaneously estimated by the group lasso, and the estimation is guided by the classification performance. Thanks to the variable selection capability of the group lasso, the framework can drop individual autoregressive coefficients that are useless in the prediction stage. Also, the framework is relatively insensitive to the chosen autoregressive order. We devise an efficient algorithm to solve this problem. We test our approach on Keirn and Aunon's data, used for binary classification of electroencephalogram signals, achieving promising results.

Neural signal classification using a simplified feature set with nonparametric clustering

This paper presents a spike sorting method using a simplified feature set with a nonparametric clustering algorithm. The proposed feature extraction algorithm is efficient and has been implemented with a custom integrated circuit chip interfaced with the PC. The proposed clustering algorithm performs nonparametric clustering. It defines an energy function to characterize the compactness of the data and proves that the clustering procedure converges. Through iterations, the data points collapse into well formed clusters and the associated energy approaches zero. By claiming these isolated clusters, neural spikes are classified.

Classification of neural signals by a generalized correlation classifier based on radial basis functions

A common problem in neuroscience is to identify the features by which a set of measurements can be segregated into different classes, for example into different responses to sensory stimuli. A main difficulty is that the derived distributions are often high-dimensional and complex. Many multivariate analysis techniques, therefore, aim to find a simpler low-dimensional representation. Most of them either involve huge efforts in implementation and data handling or ignore important structures and relationships within the original data. We developed a dimension reduction method by means of radial basis functions (RBF), where only a system of linear equations has to be solved. We show that this approach can be regarded as an extension of a linear correlation-based classifier. The validity and reliability of this technique is demonstrated on artificial data sets. Its practical relevance is further confirmed by discriminating recordings from monkey visual cortex evoked by different stimuli.

Classification of non-stationary neural signals

Although a number of methods have been proposed for classification of individual action potentials embedded in multi-unit activity, they have been challenged by non-stationarity. The waveform shapes of action potentials can change rapidly over time as a result of shifts in membrane conductances during extended burst firing sequences and more slowly over time due to electrode drift. These changes are typically non-Gaussian. We present an algorithm for waveform identification that makes no assumptions on the distribution of these shapes other than the change in waveform shape for a particular neuron should not be discontinuous. We apply this algorithm to the resolution of multi-unit neural signals recorded in the cat visual cortex and we compare this approach to a spike sorting method that is based on the Bayesian likelihood of a spike fitting a particular model (Lewicki, M. Bayesian modeling and classification of neural signals. Neural Comput 1994;6(5):1005–1030).

Bayesian Modeling and Classification of Neural Signals

Identifying and classifying action potential shapes in extracellular neural waveforms have long been the subject of research, and although several algorithms for this purpose have been successfully applied, their use has been limited by some outstanding problems. The first is how to determine shapes of the action potentials in the waveform and, second, how to decide how many shapes are distinct. A harder problem is that action potentials frequently overlap making difficult both the determination of the shapes and the classification of the spikes. In this report, a solution to each of these problems is obtained by applying Bayesian probability theory. By defining a probabilistic model of the waveform, the probability of both the form and number of spike shapes can be quantified. In addition, this framework is used to obtain an efficient algorithm for the decomposition of arbitrarily complex overlap sequences. This algorithm can extract many times more information than previous methods and facilitates the extracellular investigation of neuronal classes and of interactions within neuronal circuits.