Wavelet methods for spike detection in mouse renal sympathetic nerve activity

Spectral resonance in Fitzhugh-Nagumo neuron system: relation with stochastic resonance and its role in EMG signal characterization

This paper examines the existence of spectral resonance in the Fitzhugh-Nagumo (FHN) system driven by periodical signal and unbounded noise having Gaussian distribution. It is newly revealed that if the inter-spike-interval (ISI) distribution is accumulated on a single cluster, there exists a dual relationship between stochastic resonance and spectral resonance determined by commonly used metric normalized standard deviation of ISI. Furthermore, the ISI distribution is also concentrated on more than one cluster depending on different driving signal frequency. Consequently, the apparent regular spiking behavior is observed to occur at specified driving signal frequencies which result in a local minimum in entropy function indicating spectral resonance. Therefore it is proposed that occurrence of spectral resonance strongly depends on the shape of ISI distribution tuned by the stochastic and deterministic driving signal parameters and conventional metrics may not indicate entire resonance behavior. Correspondingly, the entropy function is utilized in this paper as an alternative metric to enable the detection of the spectral resonance occurrence. The ISI distribution obtained from the FHN system is investigated to relate the real electromyography (EMG) measurements under different conditions such as myokymia and neuromyotonia. It is seen that ISI distribution observed from myokymic EMG exhibits notably close behavior as in the case of spectral resonance generated by FHN whereas a wider distribution is monitored in the case of neuromyotonia. It is contributed that the modeling and parameterization based on ISI distribution can be potentially used to identify different neural activities.

Efficient, automatic, and optimized portable Raman-spectrum-based pesticide detection system

Raman spectroscopy can be used for accurately detecting pesticides and determining the chemical composition of a pesticide. To facilitate field detection, the present study used a portable Raman spectrometer for analysis. However, this spectrometer was found to be susceptible to noise interference and signal offsets, which increased the difficulty of pesticide identification. The most commonly used algorithm for Raman spectrum identification is principal component analysis (PCA). However, accurate classification often cannot be achieved with PCA because of the offset and noise in the Raman spectrum data. Therefore, in this study, after the collected Raman spectrum data were processed using the small-step, center-weighted moving-average method, these data were employed to train a convolutional neural network (CNN) model for prediction. To optimize the CNN model, the hyperparameters of the CNN were adjusted using various optimization algorithms, and the optimal solution was obtained after multiple iterations. Data preprocessing and architecture training models were then constructed in a self-optimized manner to improve the ability of the algorithm model to handle diverse types of data. Finally, a CNN model optimized using the cat swarm optimization algorithm was developed. This model was trained on 3000 samples containing three pesticides, and its accuracy for pesticide composition identification was discovered to be 89.33%.

NeuSort: an automatic adaptive spike sorting approach with neuromorphic models

Objective.Spike sorting, a critical step in neural data processing, aims to classify spiking events from single electrode recordings based on different waveforms. This study aims to develop a novel online spike sorter, NeuSort, using neuromorphic models, with the ability to adaptively adjust to changes in neural signals, including waveform deformations and the appearance of new neurons.Approach.NeuSort leverages a neuromorphic model to emulate template-matching processes. This model incorporates plasticity learning mechanisms inspired by biological neural systems, facilitating real-time adjustments to online parameters.Results.Experimental findings demonstrate NeuSort's ability to track neuron activities amidst waveform deformations and identify new neurons in real-time. NeuSort excels in handling non-stationary neural signals, significantly enhancing its applicability for long-term spike sorting tasks. Moreover, its implementation on neuromorphic chips guarantees ultra-low energy consumption during computation.Significance.NeuSort caters to the demand for real-time spike sorting in brain-machine interfaces through a neuromorphic approach. Its unsupervised, automated spike sorting process makes it a plug-and-play solution for online spike sorting.

EEG artifact removal using sub-space decomposition, nonlinear dynamics, stationary wavelet transform and machine learning algorithms

Blind source separation (BSS) methods have received a great deal of attention in electroencephalogram (EEG) artifact elimination as they are routine and standard signal processing tools to remove artifacts and reserve desired neural information. On the other hand, a classifier should follow BSS methods to automatically identify artifactual sources and remove them in the following steps. In addition, removing all detected artifactual components leads to loss of information since some desired information related to neural activity leaks to these sources. So, an approach should be employed to detect and suppress the artifacts and reserve neural activity. This study introduces a novel method based on EEG and Poincare planes in the phase space to detect artifactual components estimated by second-order blind identification (SOBI). Artifacts are detected using a mixture of well-known conventional classifiers and were removed employing stationary wavelet transform (SWT) to reserve neural information. The proposed method is a combination of signal processing techniques and machine learning algorithms, including multi-layer perceptron (MLP), K-nearest neighbor (KNN), naïve Bayes, and support vector machine (SVM) which have significant results while applying our proposed method to different scenarios. Simulated, semi-simulated, and real EEG signals are employed to evaluate the proposed method, and several evaluation criteria are calculated. We achieved acceptable results, for example, 98% average accuracy and 97% average sensitivity in artifactual EEG component detection or about 2% as mean square error in EEG reconstruction after artifact removal. Results showed that the proposed method is effective and can be used in future studies as we have considered different real-world scenarios to evaluate it.

An Artifact-Resistant Feature SKNAER for Quantifying the Burst of Skin Sympathetic Nerve Activity Signal

Evaluation of sympathetic nerve activity (SNA) using skin sympathetic nerve activity (SKNA) signal has attracted interest in recent studies. However, signal noises may obstruct the accurate location for the burst of SKNA, leading to the quantification error of the signal. In this study, we use the Teager−Kaiser energy (TKE) operator to preprocess the SKNA signal, and then candidates of burst areas were segmented by an envelope-based method. Since the burst of SKNA can also be discriminated by the high-frequency component in QRS complexes of electrocardiogram (ECG), a strategy was designed to reject their influence. Finally, a feature of the SKNA energy ratio (SKNAER) was proposed for quantifying the SKNA. The method was verified by both sympathetic nerve stimulation and hemodialysis experiments compared with traditional heart rate variability (HRV) and a recently developed integral skin sympathetic nerve activity (iSKNA) method. The results showed that SKNAER correlated well with HRV features (r = 0.60 with the standard deviation of NN intervals, 0.67 with low frequency/high frequency, 0.47 with very low frequency) and the average of iSKNA (r = 0.67). SKNAER improved the detection accuracy for the burst of SKNA, with 98.2% for detection rate and 91.9% for precision, inducing increases of 3.7% and 29.1% compared with iSKNA (detection rate: 94.5% (p < 0.01), precision: 62.8% (p < 0.001)). The results from the hemodialysis experiment showed that SKNAER had more significant differences than aSKNA in the long-term SNA evaluation (p < 0.001 vs. p = 0.07 in the fourth period, p < 0.01 vs. p = 0.11 in the sixth period). The newly developed feature may play an important role in continuously monitoring SNA and keeping potential for further clinical tests.

Removing Noise from Extracellular Neural Recordings Using Fully Convolutional Denoising Autoencoders

Extracellular recordings are severely contaminated by a considerable amount of noise sources, rendering the denoising process an extremely challenging task that should be tackled for efficient spike sorting. To this end, we propose an end-to-end deep learning approach to the problem, utilizing a Fully Convolutional Denoising Autoencoder, which learns to produce a clean neuronal activity signal from a noisy multichannel input. The experimental results on simulated data show that our proposed method can improve significantly the quality of noise-corrupted neural signals, outperforming widely-used wavelet denoising techniques.

Capturing spike train temporal pattern with wavelet average coefficient for brain machine interface

Motor brain machine interfaces (BMIs) directly link the brain to artificial actuators and have the potential to mitigate severe body paralysis caused by neurological injury or disease. Most BMI systems involve a decoder that analyzes neural spike counts to infer movement intent. However, many classical BMI decoders (1) fail to take advantage of temporal patterns of spike trains, possibly over long time horizons; (2) are insufficient to achieve good BMI performance at high temporal resolution, as the underlying Gaussian assumption of decoders based on spike counts is violated. Here, we propose a new statistical feature that represents temporal patterns or temporal codes of spike events with richer description-wavelet average coefficients (WAC)-to be used as decoder input instead of spike counts. We constructed a wavelet decoder framework by using WAC features with a sliding-window approach, and compared the resulting decoder against classical decoders (Wiener and Kalman family) and new deep learning based decoders ( Long Short-Term Memory) using spike count features. We found that the sliding-window approach boosts decoding temporal resolution, and using WAC features significantly improves decoding performance over using spike count features.

Systematic analysis of wavelet denoising methods for neural signal processing

Objective.Among the different approaches for denoising neural signals, wavelet-based methods are widely used due to their ability to reduce in-band noise. All wavelet denoising algorithms have a common structure, but their effectiveness strongly depends on several implementation choices, including the mother wavelet, the decomposition level, the threshold definition, and the way it is applied (i.e. the thresholding). In this work, we investigated these factors to quantitatively assess their effects on neural signals in terms of noise reduction and morphology preservation, which are important when spike sorting is required downstream.Approach.Based on the spectral characteristics of the neural signal, according to the sampling rate of the signals, we considered two possible decomposition levels and identified the best-performing mother wavelet. Then, we compared different threshold estimation and thresholding methods and, for the best ones, we also evaluated their effect on clearing the approximation coefficients. The assessments were performed on synthetic signals that had been corrupted by different types of noise and on a murine peripheral nervous system dataset, both of which were sampled at about 16 kHz. The results were statistically analysed in terms of their Pearson's correlation coefficients, root-mean-square errors, and signal-to-noise ratios.Main results.As expected, the wavelet implementation choices greatly influenced the processing performance. Overall, the Haar wavelet with a five-level decomposition, hard thresholding method, and the threshold proposed by Hanet al(2007) achieved the best outcomes. Based on the adopted performance metrics, wavelet denoising with these parametrizations outperformed conventional 300-3000 Hz linear bandpass filtering.Significance.These results can be used to guide the reasoned and accurate selection of wavelet denoising implementation choices in the context of neural signal processing, particularly when spike-morphology preservation is required.

Nondestructive Evaluation of Thermal Barrier Coatings Interface Delamination Using Terahertz Technique Combined with SWT-PCA-GA-BP Algorithm

Thermal barrier coatings (TBCs) are usually subjected to the combined action of compressive stress, tensile stress, and bending shear stress, resulting in the interfacial delamination of TBCs, and finally causing the ceramic top coat to peel off. Hence, it is vital to detect the early-stage subcritical delamination cracks. In this study, a novel hybrid artificial neural network combined with the terahertz nondestructive technology was presented to predict the thickness of interface delamination in the early stage. The finite difference time domain (FDTD) algorithm was used to obtain the raw terahertz time-domain signals of 32 TBCs samples with various thicknesses of interface delamination, not only that, the influence of roughness and the thickness of the ceramic top layer were considered comprehensively when modeling. The stationary wavelet transform (SWT) and principal component analysis (PCA) methods were employed to extract the signal features and reduce the data dimensions before modeling, to make the cumulative contribution rate reach 100%, the first 31 components of the SWT detail data was used as the input data during modeling. Finally, a back propagation (BP) neural network method optimized by the genetic algorithm (GA-BP) was proposed to set up the interface delamination thickness prediction model. As a result, the root correlation coefficient R2 reached over 0.95, the various errors—including the mean square error, mean squared percentage error, and mean absolute percentage error—were less than or equal to 0.53. All these indicators proved that the trained hybrid SWT-PCA-GA-BP model had excellent prediction performance and high accuracy. Finally, this work proposed a novel and convenient interface delamination evaluation method that could also be potentially utilized to evaluate the structural integrity of TBCs.

Extracting wavelet based neural features from human intracortical recordings for neuroprosthetics applications

Background

Understanding the long-term behavior of intracortically-recorded signals is essential for improving the performance of Brain Computer Interfaces. However, few studies have systematically investigated chronic neural recordings from an implanted microelectrode array in the human brain.

Methods

In this study, we show the applicability of wavelet decomposition method to extract and demonstrate the utility of long-term stable features in neural signals obtained from a microelectrode array implanted in the motor cortex of a human with tetraplegia. Wavelet decomposition was applied to the raw voltage data to generate mean wavelet power (MWP) features, which were further divided into three sub-frequency bands, low-frequency MWP (lf-MWP, 0–234 Hz), mid-frequency MWP (mf-MWP, 234 Hz–3.75 kHz) and high-frequency MWP (hf-MWP, >3.75 kHz). We analyzed these features using data collected from two experiments that were repeated over the course of about 3 years and compared their signal stability and decoding performance with the more standard threshold crossings, local field potentials (LFP), multi-unit activity (MUA) features obtained from the raw voltage recordings.

Results

All neural features could stably track neural information for over 3 years post-implantation and were less prone to signal degradation compared to threshold crossings. Furthermore, when used as an input to support vector machine based decoding algorithms, the mf-MWP and MUA demonstrated significantly better performance, respectively, in classifying imagined motor tasks than using the lf-MWP, hf-MWP, LFP, or threshold crossings.

Conclusions

Our results suggest that using MWP features in the appropriate frequency bands can provide an effective neural feature for brain computer interface intended for chronic applications.

Trial registration

This study was approved by the U.S. Food and Drug Administration (Investigational Device Exemption) and the Ohio State University Medical Center Institutional Review Board (Columbus, Ohio). The study conformed to institutional requirements for the conduct of human subjects and was filed on ClinicalTrials.gov (Identifier NCT01997125).

Electronic supplementary material

The online version of this article (10.1186/s42234-018-0011-x) contains supplementary material, which is available to authorized users.

Spike detection using a multiresolution entropy based method

Correct interpretation of neural mechanisms depends on the accurate detection of neuronal activities, which become visible as spikes in the electrical activity of neurons. In the present work, a novel entropy based method is proposed for spike detection which employs the fact that transient spike events change the entropy level of the neural time series. In this regard, the time-dependent entropy method can be used for detecting spike times, where the entropy of a selected segment of a neural time series, using a sliding window approach, is calculated and the time of the events are highlighted by sharp peaks in the output of the time-dependent entropy method. It is shown that the length of the sliding window determines the resolution of the time series in entropy space, therefore, the calculation is performed with a different window length for obtaining a multiresolution transform. The final decision threshold for detecting spike events is applied to the point-wise product of the time dependent entropy calculations with different resolutions. The proposed detection method has been assessed using several simulated and real neural data sets. The results show that the proposed method detects spikes in their exact times while compared with other traditional methods, relatively lower false alarm rate is obtained.

Spike rate of multi-unit muscle sympathetic nerve fibers after catheter-based renal nerve ablation

Patients with treatment-resistant arterial hypertension exhibited profound reductions in single sympathetic vasoconstrictor fiber firing rates after renal nerve ablation. In contrast, integrated multi-unit muscle sympathetic nerve activity (MSNA) changed little or not at all. We hypothesized that conventional MSNA analysis may have missed single fiber discharges, thus, obscuring sympathetic inhibition after renal denervation. We studied patients with difficult-to-control arterial hypertension (age 45-74 years) before, 6 (n = 11), and 12 months (n = 8) after renal nerve ablation. Electrocardiogram, respiration, brachial, and finger arterial blood pressure (BP), as well as the MSNA and raw MSNA signals were analyzed. We detected MSNA action-potential spikes using 2 stage kurtosis wavelet denoising techniques to assess mean, median, and maximum spike rates for each beat-to-beat interval. Supine heart rate and systolic BP did not change at 6 (ΔHR: -2 ± 3 bpm; ΔSBP: 2 ± 9 mm Hg) or at 12 months (ΔHR: -1 ± 3 mm Hg, ΔSBP: -1 ± 9 mm Hg) after renal nerve ablation. Mean burst frequency and mean spike frequency at baseline were 34 ± 3 bursts per minute and 8 ± 1 spikes per second. Both measurements did not change at 6 months (-1.4 ± 3.6 bursts/minute; -0.6 ± 1.4 spikes/second) or at 12 months (-2.5 ± 4.0 bursts/minute; -2.0 ± 1.6 spikes/second) after renal nerve ablation. After renal nerve ablation, BP decreased in 3 of 11 patients. BP and MSNA spike frequency changes were not correlated (slope = -0.06; P = .369). Spike rate analysis of multi-unit MSNA neurograms further suggests that profound sympathetic inhibition is not a consistent finding after renal nerve ablation.

A supervised multi-sensor matched filter for the detection of extracellular action potentials

Multi-sensor extracellular recording takes advantage of several electrode channels to record from multiple neurons at the same time. However, the resulting low signal-to-noise ratio (SNR) combined with biological noise makes signal detection, the first step of any neurophysiological data analysis, difficult. A matched filter was therefore designed to better detect extracellular action potentials (EAPs) from multi-sensor extracellular recordings. The detector was tested on tetrode data from a locust antennal lobe and assessed against three trained analysts. 25 EAPs and noise samples were selected manually from the data and used for training. To reduce complexity, the filter assumed that the underlying noise in the data was spatially white. The detector performed with an average TP and FP rate of 84.62% and 16.63% respectively. This high level of performance indicates the algorithm is suitable for widespread use.

To sort or not to sort: the impact of spike-sorting on neural decoding performance

OBJECTIVE

Brain-computer interfaces (BCIs) are a promising technology for restoring motor ability to paralyzed patients. Spiking-based BCIs have successfully been used in clinical trials to control multi-degree-of-freedom robotic devices. Current implementations of these devices require a lengthy spike-sorting step, which is an obstacle to moving this technology from the lab to the clinic. A viable alternative is to avoid spike-sorting, treating all threshold crossings of the voltage waveform on an electrode as coming from one putative neuron. It is not known, however, how much decoding information might be lost by ignoring spike identity.

APPROACH

We present a full analysis of the effects of spike-sorting schemes on decoding performance. Specifically, we compare how well two common decoders, the optimal linear estimator and the Kalman filter, reconstruct the arm movements of non-human primates performing reaching tasks, when receiving input from various sorting schemes. The schemes we tested included: using threshold crossings without spike-sorting; expert-sorting discarding the noise; expert-sorting, including the noise as if it were another neuron; and automatic spike-sorting using waveform features. We also decoded from a joint statistical model for the waveforms and tuning curves, which does not involve an explicit spike-sorting step.

MAIN RESULTS

Discarding the threshold crossings that cannot be assigned to neurons degrades decoding: no spikes should be discarded. Decoding based on spike-sorted units outperforms decoding based on electrodes voltage crossings: spike-sorting is useful. The four waveform based spike-sorting methods tested here yield similar decoding efficiencies: a fast and simple method is competitive. Decoding using the joint waveform and tuning model shows promise but is not consistently superior.

SIGNIFICANCE

Our results indicate that simple automated spike-sorting performs as well as the more computationally or manually intensive methods used here. Even basic spike-sorting adds value to the low-threshold waveform-crossing methods often employed in BCI decoding.

Spike detection based on normalized correlation with automatic template generation

A novel feedback-based spike detection algorithm for noisy spike trains is presented in this paper. It uses the information extracted from the results of spike classification for the enhancement of spike detection. The algorithm performs template matching for spike detection by a normalized correlator. The detected spikes are then sorted by the OSortalgorithm. The mean of spikes of each cluster produced by the OSort algorithm is used as the template of the normalized correlator for subsequent detection. The automatic generation and updating of templates enhance the robustness of the spike detection to input trains with various spike waveforms and noise levels. Experimental results show that the proposed algorithm operating in conjunction with OSort is an efficient design for attaining high detection and classification accuracy for spike sorting.

Data compression in brain-machine/computer interfaces based on the Walsh-Hadamard transform

This paper reports on the application of the Walsh-Hadamard transform (WHT) for data compression in brain-machine/brain-computer interfaces. Using the proposed technique, the amount of the neural data transmitted off the implant is compressed by a factor of at least 63 at the expense of as low as 4.66% RMS error between the signal reconstructed on the external host and the original neural signal on the implant side. Based on the proposed idea, a 128-channel WHT processor was designed in a 0.18- μm CMOS process occupying 1.64 mm(2) of silicon area. The circuit consumes 81 μW (0.63 μW per channel) from a 1.8-V power supply at 250 kHz. A prototype of the proposed processor was implemented and successfully tested using prerecorded neural signals.

Electrically elicited visual evoked potentials in Argus II retinal implant wearers

PURPOSE

We characterized electrically elicited visual evoked potentials (eVEPs) in Argus II retinal implant wearers.

METHODS

eVEPs were recorded in four subjects, and analyzed by determining amplitude and latency of the first two positive peaks (P1 and P2). Subjects provided subjective feedback by rating the brightness and size of the phosphenes. We established eVEP input-output relationships, eVEP variability between and within subjects, the effect of stimulating different areas of the retina, and the maximal pulse rate to record eVEPs reliably.

RESULTS

eVEP waveforms had low signal-to-noise ratios, requiring long recording times and substantial signal processing. Waveforms varied between subjects, but showed good reproducibility and consistent parameter dependence within subjects. P2 amplitude overall was the most robust outcome measure and proved an accurate indicator of subjective threshold. Peak latencies showed small within-subject variability, yet their correlation with stimulus level and subjective rating were more variable than that of peak amplitudes. Pulse rates of up to (2)/3 Hz resulted in reliable eVEP recordings. Perceived phosphene brightness declined over time, as reflected in P1 amplitude, but not in P2 amplitude or peak latencies. Stimulating-electrode location significantly affected P1 and P2 amplitude and latency, but not subjective percepts.

CONCLUSIONS

While recording times and signal processing are more demanding than for standard visually evoked potential (VEP) recordings, the eVEP has proven to be a reliable tool to verify retinal implant functionality. eVEPs correlated with various stimulus parameters and with perceptual ratings. In view of these findings, eVEPs may become an important tool in functional investigations of retinal prostheses. (ClinicalTrials.gov number NCT00407602.) Dutch Abstract.

Noise suppressed, multifocus image fusion for enhanced intraoperative navigation

Current intraoperative imaging systems are typically not able to provide 'sharp' images over entire large areas or entire organs. Distinct structures such as tissue margins or groups of malignant cells are therefore often difficult to detect, especially under low signal-to-noise-ratio conditions. In this report, we introduce a noise suppressed multifocus image fusion algorithm, that provides detailed reconstructions even when images are acquired under sub-optimal conditions, such is the case for real time fluorescence intraoperative surgery. The algorithm makes use of the Anscombe transform combined with a multi-level stationary wavelet transform with individual threshold-based shrinkage. While the imaging system is integrated with a respiratory monitor triggering system, it can be easily adapted to any commercial imaging system. The developed algorithm is made available as a plugin for Osirix.

An unsupervised method for on-chip neural spike detection in multi-electrode recording systems

Emerging multi-electrode-based brain-machine interfaces (BMIs) and large multi-electrode arrays used in in vitro experiments, enable recording of single neuron's activity on multiple electrodes and allow for an in-depth investigation of neural preparations, even at a sub-cellular level. However, the use of these devices entails stringent area and power consumption constraints for the signal-processing hardware units. In addition, the high autonomy of these units and an ability to automatically adapt to changes in the recorded neural preparations is required. Implementing spike detection in close proximity to recording electrodes offers the advantage of reducing the transmission data bandwidth. By eliminating the need of transmitting the full, redundant recordings of neural activity and by transmitting only the spike waveforms or spike times, significant power savings can be achieved in the majority of cases. Here, we present a low-complexity, unsupervised, adaptable, real-time spike-detection method targeting multi-electrode recording devices and compare this method to other spike-detection methods with regard to complexity and performance.

Stationary wavelet transform and higher order statistical analyses of intrafascicular nerve recordings

Nerve signals were recorded from the sciatic nerve of the rabbits in the acute experiments with multi-channel thin-film longitudinal intrafascicular electrodes. 5.5 s sequences of quiescent and high-level nerve activity were spectrally decomposed by applying a ten-level stationary wavelet transform with the Daubechies 10 (Db10) mother wavelet. Then, the statistical distributions of the raw and subband-decomposed sequences were estimated and used to fit a fourth-order Pearson distribution as well as check for normality. The results indicated that the raw and decomposed background and high-level nerve activity distributions were nearly zero-mean and non-skew. All distributions with the frequency content above 187.5 Hz were leptokurtic except for the first-level decomposition representing frequencies in the subband between 12 and 24 kHz, which was Gaussian. This suggests that nerve activity acts to change the statistical distribution of the recording. The results further demonstrated that quiescent recording contained a mixture of an underlying pink noise and low-level nerve activity that could not be silenced. The signal-to-noise ratios based upon the standard deviation (SD) and kurtosis were estimated, and the latter was found as an effective measure for monitoring the nerve activity residing in different frequency subbands. The nerve activity modulated kurtosis along with SD, suggesting that the joint use of SD and kurtosis could improve the stability and detection accuracy of spike-detection algorithms. Finally, synthesizing the reconstructed subband signals following denoising based upon the higher order statistics of the subband-decomposed coefficient sequences allowed us to effectively purify the signal without distorting spike shape.

Control and data acquisition software for high-density CMOS-based microprobe arrays implementing electronic depth control

This paper presents the NeuroSelect software for managing the electronic depth control of cerebral CMOS-based microprobes for extracellular in vivo recordings. These microprobes contain up to 500 electronically switchable electrodes which can be appropriately selected with regard to specific neuron locations in the course of a recording experiment. NeuroSelect makes it possible to scan the electrodes electronically and to (re)select those electrodes of best signal quality resulting in a closed-loop design of a neural acquisition system. The signal quality is calculated by the relative power of the spikes compared with the background noise. The spikes are detected by an adaptive threshold using a robust estimator of the standard deviation. Electrodes can be selected in a manual or semi-automatic mode based on the signal quality. This electronic depth control constitutes a significant improvement for multielectrode probes, given that so far the only alternative has been the fine positioning by mechanical probe translation. In addition to managing communication with the hardware controller of the probe array, the software also controls acquisition, processing, display and storage of the neural signals for further analysis.

Combination of PCA and undecimated wavelet transform for neural data processing

Nervous system conveys information by electrical signals called 'spikes', therefore, spikes detection and sorting are challenging topics in the neural data processing. The principal component analysis (PCA) is a convenient tool for clustering spikes; however it has some disadvantages for closely shaped and overlapped spikes. For such the cases, an algorithm based on the combination of the principal component analysis and undecimated wavelet transform, is proposed to enhance the cluster formation from the spikes mapping. These results indicate that the principal component analysis used in combination with the undecimated wavelet has a better performance in the spike sorting. This can lead to more compact and separate clusters in comparison with the PCA clustering and more efficient spike sorting.

Spike detection in human muscle sympathetic nerve activity using a matched wavelet approach

Sympathetic nerve recordings associated with blood pressure regulation can be recorded directly using microneurography. A general characteristic of this signal is spontaneous burst activity of spikes (action potentials) separated by silent periods against a background of considerable Gaussian noise. During measurement with electrodes, the raw muscle sympathetic nerve activity (MSNA) signal is amplified, band-pass filtered, rectified and integrated. This integration process removes information regarding action potential content and their discharge properties. This paper proposes a new method for detecting action potentials from the raw MSNA signal to enable investigation of post-ganglionic neural discharge properties. The new method is based on the design of a mother wavelet that is matched to an actual mean action potential template extracted from a real raw MSNA signal. To detect action potentials, the new matched wavelet is applied to the MSNA signal using a continuous wavelet transform following a thresholding procedure and finding of a local maxima that indicates the location of action potentials. The performance of the proposed method versus two previous wavelet-based approaches was evaluated using (1) real MSNA recorded from seven healthy participants and, (2) simulated MSNA. The results show that the new matched wavelet performs better than the previous wavelet-based methods that use a non-matched wavelet in detecting action potentials in the MSNA signal.

Quantifying time-varying multiunit neural activity using entropy based measures

Modern micro-electrode arrays make it possible to simultaneously record population neural activity. However, methods to analyze multiunit activity (MUA), which reflects the aggregate spiking activity of a population of neurons, have remained underdeveloped in comparison to those used for studying single unit activity (SUA). In scenarios where SUA is hard to record and maintain or is not representative of brains response, MUA is informative in deciphering the brains complex time-varying response to stimuli or to clinical insults. Here, we present two quantitative methods of analysis of the time-varying dynamics of MUA without spike detection. These methods are based on the multiresolution discrete wavelet transform (DWT) of an envelope of MUA followed by information theoretic measures: multiresolution entropy (MRE) and the multiresolution Kullback-Leibler distance (MRKLD). We test the proposed quantifiers on both simulated and experimental MUA recorded from rodent cortex in an experimental model of global hypoxic-ischemic brain injury. First, our results validate the use of the envelope of MUA as an alternative to detecting and analyzing transient and complex spike activity. Second, the MRE and MRKLD are shown to respond to dynamic changes due to the brains response to global injury and to identify the transient changes in the MUA.

Multiresolution entropy measure for neuronal multiunit activity

It is known that the multiunit activity (MUA) reflects the status of population of neurons in the vicinity of an electrode. We provide a quantitative measure of the time-varying multiunit neuronal spiking activity using an entropy based approach. To verify the status embedded in the neuronal activity of a population of neurons, we incorporate the discrete wavelet transform (DWT) to isolate the inherent spiking activity of MUA from the noise and background cortical activity or field potentials. Owing to the decorrelating property of DWT, the spiking activity would be preserved while reducing the non-spiking component such as the background noise. By evaluating the entropy of the wavelet coefficients of the denoised MUA, a multiresolution entropy of the MUA signal is developed. The proposed entropy measure was tested in the analysis of both simulated noisy MUA and actual MUA recorded from cortex in rodent model which undergoes hypoxic-ischemic brain injury. Simulation and Experimental results demonstrate that the dynamics of a population can be quantified by using the proposed multiresolution entropy.

Detection of multifiber neuronal firings: a mixture separation model applied to sympathetic recordings

Sympathetic nervous flow to the vasculature plays a critical role in control of regional blood flow; however, traditional processing methods of multifiber recordings cannot reliably discriminate physiologically irrelevant information from actual nerve activity, and alternative wavelet methods suffer from subjectivity and lack of a well-specified model. We propose an algorithm that allows objective threshold selection under general assumptions regarding the sparsity and statistical structure of the neural signal and noise. Our study shows that the conditional expectation of the actual nerve signal can be estimated and used to maximize the signal-to-noise ratio (SNR). We evaluated the algorithm's performance on artificial datasets and on actual multifiber recordings (44 datasets from 22 subjects, and 1 set from a rat). On artificial datasets, the algorithm identified 70% and 80% of the spikes at -3.5 and 0.5 dB SNR with a good match between the actual and estimated spike count (R2 = 0.719, p < 0.001). On actual recordings, the overall improvement in performance compared to that of a traditional processing method was significant (t = 3.88; p = 0.0002). Our results show the applicability of this algorithm to multifiber recordings not only in humans, but also in other species.

Detection and classification of raw action potential patterns in human Muscle Sympathetic Nerve Activity

The Muscle Sympathetic Nerve Activity (MSNA) consists of synchronous neural discharges separated by periods of neural silence dominated by heavy background noise. During measurement with electrodes, the raw MSNA signal is amplified, band-pass filtered, rectified and integrated. This integration process removes much neurophysiological information. In this paper a method for detecting a raw action potential (before the pre-amplifier) and filtered action potential (after the band-pass filter) is presented. This method is based on stationary wavelet transform (SWT) and a peak detection algorithm. Also, the detected action potentials were clustered using the k-means method and the cluster averages were calculated. The action potential detector and classification algorithm are evaluated using real MSNA recorded from three healthy subjects.

Performance analysis of stationary and discrete wavelet transform for action potential detection from sympathetic nerve recordings in humans

Accurate investigation of the sympathetic nervous system is important in the diagnosis and study of various autonomic and cardiovascular control and disorders. Sympathetic function associated with blood pressure regulation in humans can be evaluated by recording muscle sympathetic nerve activity (MSNA), which is characterised by synchronous neuronal discharges separated by periods of neural silence dominated by colored gaussian noise. In this paper two common methods for detecting filtered action potential in MSNA recordings is compared. These methods are based on stationary wavelet transform (SWT) and discrete wavelet transform (DWT). The performance analysis are evaluated using simulated MSNA using templates extracted from real MSNA recorded from three healthy subjects.

Optimizing the automatic selection of spike detection thresholds using a multiple of the noise level

Thresholding is an often-used method of spike detection for implantable neural signal processors due to its computational simplicity. A means for automatically selecting the threshold is desirable, especially for high channel count data acquisition systems. Estimating the noise level and setting the threshold to a multiple of this level is a computationally simple means of automatically selecting a threshold. We present an analysis of this method as it is commonly applied to neural waveforms. Four different operators were used to estimate the noise level in neural waveforms and set thresholds for spike detection. An optimal multiplier was identified for each noise measure using a metric appropriate for a brain-machine interface application. The commonly used root-mean-square operator was found to be least advantageous for setting the threshold. Investigators using this form of automatic threshold selection or developing new unsupervised methods can benefit from the optimization framework presented here.

Wavelet-based scale-dependent detection of neurological action potentials

We study different wavelet-based algorithms for the detection of neurological action potentials recorded using micro-electrode arrays (MEA). We plan to develop a new family of ASIC-embedded low power algorithms close to the recording sites. We use the wavelet theory, not for previous-to-the-detection denoising stage (as it is usually used for) but for the detection itself. Different adaptive methods are presented with varying complexity levels. We demonstrate that wavelet-based detection of extracellular action potentials is superior than traditional and simpler approaches, at the expense of a slightly larger computational load. Moreover, our method is shown to be fully compatible with an embedded implementation. Proposed algorithms are applied to simulated datasets using a simplified model of the American cockroach antennal lobe.

Comparison of spike-sorting algorithms for future hardware implementation

Applications such as brain-machine interfaces require hardware spike sorting in order to (1) obtain single-unit activity and (2) perform data reduction for wireless transmission of data. Such systems must be low-power, low-area, high-accuracy, automatic, and able to operate in real time. Several detection and feature extraction algorithms for spike sorting are described briefly and evaluated in terms of accuracy versus computational complexity. The nonlinear energy operator method is chosen as the optimal spike detection algorithm, being most robust over noise and relatively simple. The discrete derivatives method [1] is chosen as the optimal feature extraction method, maintaining high accuracy across SNRs with a complexity orders of magnitude less than that of traditional methods such as PCA.

Sympathetic nerve traffic and circulating norepinephrine levels in RGS2-deficient mice

Regulator of G protein signaling 2 (RGS2-/-) deficient mice feature an increased resting blood pressure and an excessive pressor response to stress. We measured renal sympathetic nerve activity (RSNA) directly to test the hypothesis that RSNA is increased in RGS2-/- mice, compared to RGS2+/+ mice. Seventeen mice (RGS2-/-, n=9; RGS2+/+, n=8) were anesthetized with isoflurane. We cannulated the left jugular vein for drug administration. Renal sympathetic nerve activity (RSNA) was recorded using bipolar electrodes. Arterial blood pressure (BP) from the femoral artery, ECG (needle electrodes), and RSNA were recorded (sample rate 10 kHz) simultaneously. RSNA was analysed off-line using a modified wavelet de-noising technique and the classical discriminator method. RSNA detected during phenylephrine bolus injections or after the animals death was subtracted from baseline values. Mean arterial blood pressure, norepinephrine plasma levels, the responsiveness to vasoactive drugs, and the sympathetic baroreflex gain were similar in anesthetized RGS2+/+ and RGS2-/- animals. RSNA was lower in RGS2-/- mice compared to wild-type controls (wavelet: spike rate in Hz: RGS2+/+ 25.5+/-5.1; RGS2-/- 17.4+/-4.0; discriminator method: RGS2+/+ 41.4+/-5.7, RGS2-/- 22.0+/-4.3, p<0.05). Thus, the expected result proved not to be the case. Our data suggest a mismatch between sympathetic nerve traffic and plasma norepinephrine concentrations. This observation may depend on altered coupling between electrical nerve activity and norepinephrine release and/or a changed norepinephrine uptake in RGS2-/- mice.

A simplified two-component model of blood pressure fluctuation

We propose a simple moving-average (MA) model that uses the low-frequency (LF) component of the peroneal muscle sympathetic nerve spike rate (LF(spike rate)) and the high-frequency (HF) component of respiration (HF(Resp)) to describe the LF neurovascular fluctuations and the HF mechanical oscillations in systolic blood pressure (SBP), respectively. This method was validated by data from eight healthy subjects (23-47 yr old, 6 male, 2 female) during a graded tilt (15 degrees increments every 5 min to a 60 degrees angle). The LF component of SBP (LF(SBP)) had a strong baroreflex-mediated feedback correlation with LF(spike rate) (r = -0.69 +/- 0.05) and also a strong feedforward relation to LF(spike rate) [r = 0.58 +/- 0.03 with LF(SBP) delay (tau) = 5.625 +/- 0.15 s]. The HF components of spike rate (HF(spike rate)) and SBP (HF(SBP)) were not significantly correlated. Conversely, HF(Resp) and HF(SBP) were highly correlated (r = -0.79 +/- 0.04), whereas LF(Resp) and LF(SBP) were significantly less correlated (r = 0.45 +/- 0.08). The mean correlation coefficients between the measured and model-predicted LF(SBP) (r = 0.74 +/- 0.03) in the supine position did not change significantly during tilt. The mean correlation between the measured and model-predicted HF(SBP) was 0.89 +/- 0.02 in the supine position. R(2) values for the regression analysis of the model-predicted and measured LF and HF powers indicate that 78 and 91% of the variability in power can be explained by the linear relation of LF(spike rate) to LF(SBP) and HF(Resp) to HF(SBP). We report a simple two-component model using neural sympathetic and mechanical respiratory inputs that can explain the majority of blood pressure fluctuation at rest and during orthostatic stress in healthy subjects.