Artificial neural network (ANN)‐based prediction of depth filter loading capacity for filter sizing

This article presents an application of artificial neural network (ANN) modelling towards prediction of depth filter loading capacity for clarification of a monoclonal antibody (mAb) product during commercial manufacturing. The effect of operating parameters on filter loading capacity was evaluated based on the analysis of change in the differential pressure (DP) as a function of time. The proposed ANN model uses inlet stream properties (feed turbidity, feed cell count, feed cell viability), flux, and time to predict the corresponding DP. The ANN contained a single output layer with ten neurons in hidden layer and employed a sigmoidal activation function. This network was trained with 174 training points, 37 validation points, and 37 test points. Further, a pressure cut‐off of 1.1 bar was used for sizing the filter area required under each operating condition. The modelling results showed that there was excellent agreement between the predicted and experimental data with a regression coefficient (R²) of 0.98. The developed ANN model was used for performing variable depth filter sizing for different clarification lots. Monte‐Carlo simulation was performed to estimate the cost savings by using different filter areas for different clarification lots rather than using the same filter area. A 10% saving in cost of goods was obtained for this operation. © 2016 American Institute of Chemical Engineers Biotechnol. Prog., 32:1436–1443, 2016     

A Novel Preprocessing Method Using Hilbert Huang Transform for MALDI-TOF and SELDI-TOF Mass Spectrometry Data

Motivation

Mass spectrometry is a high throughput, fast, and accurate method of protein analysis. Using the peaks detected in spectra, we can compare a normal group with a disease group. However, the spectrum is complicated by scale shifting and is also full of noise. Such shifting makes the spectra non-stationary and need to align before comparison. Consequently, the preprocessing of the mass data plays an important role during the analysis process. Noises in mass spectrometry data come in lots of different aspects and frequencies. A powerful data preprocessing method is needed for removing large amount of noises in mass spectrometry data.

Results

Hilbert-Huang Transformation is a non-stationary transformation used in signal processing. We provide a novel algorithm for preprocessing that can deal with MALDI and SELDI spectra. We use the Hilbert-Huang Transformation to decompose the spectrum and filter-out the very high frequencies and very low frequencies signal. We think the noise in mass spectrometry comes from many sources and some of the noises can be removed by analysis of signal frequence domain. Since the protein in the spectrum is expected to be a unique peak, its frequence domain should be in the middle part of frequence domain and will not be removed. The results show that HHT, when used for preprocessing, is generally better than other preprocessing methods. The approach not only is able to detect peaks successfully, but HHT has the advantage of denoising spectra efficiently, especially when the data is complex. The drawback of HHT is that this approach takes much longer for the processing than the wavlet and traditional methods. However, the processing time is still manageable and is worth the wait to obtain high quality data.     

Linear model of brain electrical activity—EEG as a superposition of damped oscillatory modes

EEG time series were modeled as an output of the linear filter driven by white noise. Parameters describing the signal were determined in a way fulfilling the maximum entropy principle. Transfer function and the impulse response function were found. The solutions of the differential equations describing the system have the form of the damped oscillatory modes. The representation of the EEG time series as a superposition of the resonant modes with characteristic decay factors seems a valuable method of the analysis of the signal, since it offers high reduction of the data to the few parameters of a clear physiological meaning.     

Time-frequency analysis and filtering of kinematic signals with impacts using the Wigner function: accurate estimation of the second derivative

Biomechanical signals are represented in the time-frequency domain using the Wigner distribution function. Filtering of this representation for the case of a non-stationary displacement signal with impact is studied. Smoothed displacement data are then double differentiated and compared with references accelerometer data. It is shown that this technique is able to remove noise from these signals in a better way than conventional filtering techniques currently used in biomechanics.     

A Flattest Constrained Envelope Approach for Empirical Mode Decomposition

Empirical mode decomposition (EMD) is an adaptive method for nonlinear, non-stationary signal analysis. However, the upper and lower envelopes fitted by cubic spline interpolation (CSI) may often occur overshoots. In this paper, a new envelope fitting method based on the flattest constrained interpolation is proposed. The proposed method effectively integrates the difference between extremes into the cost function, and applies a chaos particle swarm optimization method to optimize the derivatives of the interpolation nodes. The proposed method was tested on three different types of data: ascertain signal, random signals and real electrocardiogram signals. The experimental results show that: (1) The proposed flattest envelope effectively solves the overshoots caused by CSI method and the artificial bends caused by piecewise parabola interpolation (PPI) method. (2) The index of orthogonality of the intrinsic mode functions (IMFs) based on the proposed method is 0.04054, 0.02222±0.01468 and 0.04013±0.03953 for the ascertain signal, random signals and electrocardiogram signals, respectively, which is lower than the CSI method and the PPI method, and means the IMFs are more orthogonal. (3) The index of energy conversation of the IMFs based on the proposed method is 0.96193, 0.93501±0.03290 and 0.93041±0.00429 for the ascertain signal, random signals and electrocardiogram signals, respectively, which is closer to 1 than the other two methods and indicates the total energy deviation amongst the components is smaller. (4) The comparisons of the Hilbert spectrums show that the proposed method overcomes the mode mixing problems very well, and make the instantaneous frequency more physically meaningful.     

基于虚拟仪器技术的血管壁动态信息无创检测与辅助诊断系统

为了在体外无损地实现对血管壁动态信息、心电和心音信号等医学信息多参数的综合同步检测以及分析和处理,利用虚拟仪器技术设计了血管壁动态信息多参数的无创检测辅助诊断系统．该系统硬件平台由信号输入模块、信号调理模块以及采集卡等三大模块组成。软件系统由开发虚拟仪器的流行软件LabVIEW编写,实现了数据的采集、实时显示、分析处理和存储等。初步的临床检测结果证实了本无创检测系统的可行性和临床应用的前景,为功能性无创辅助诊断与血管壁弹性（硬化）程度有关的血管疾病提供了一种新的方法。     

Real-time spectrum estimation–based dual-channel speech-enhancement algorithm for cochlear implant

Background

Improvement of the cochlear implant (CI) front-end signal acquisition is needed to increase speech recognition in noisy environments. To suppress the directional noise, we introduce a speech-enhancement algorithm based on microphone array beamforming and spectral estimation. The experimental results indicate that this method is robust to directional mobile noise and strongly enhances the desired speech, thereby improving the performance of CI devices in a noisy environment.

Methods

The spectrum estimation and the array beamforming methods were combined to suppress the ambient noise. The directivity coefficient was estimated in the noise-only intervals, and was updated to fit for the mobile noise.

Results

The proposed algorithm was realized in the CI speech strategy. For actual parameters, we use Maxflat filter to obtain fractional sampling points and cepstrum method to differentiate the desired speech frame and the noise frame. The broadband adjustment coefficients were added to compensate the energy loss in the low frequency band.

Discussions

The approximation of the directivity coefficient is tested and the errors are discussed. We also analyze the algorithm constraint for noise estimation and distortion in CI processing. The performance of the proposed algorithm is analyzed and further be compared with other prevalent methods.

Conclusions

The hardware platform was constructed for the experiments. The speech-enhancement results showed that our algorithm can suppresses the non-stationary noise with high SNR. Excellent performance of the proposed algorithm was obtained in the speech enhancement experiments and mobile testing. And signal distortion results indicate that this algorithm is robust with high SNR improvement and low speech distortion.     

Dynamic Digital Channelizer Based on Spectrum Sensing

The ability to efficiently channelize a received signal with dynamic sub-channel bandwidths is a key requirement of software defined radio (SDR) systems. The digital channelizer, which is used to split the received signal into a number of sub-channels, plays an important role in SDR systems. In this paper, a design of dynamic digital channelizer is presented. The proposed method is novel in that it employs a cosine modulated filter bank (CMFB) to divide the received signal into multiple frequency sub-bands and a spectrum sensing technique, which is mostly used in cognitive radio, is introduced to detect the presence of signal of each sub-band. The method of spectrum sensing is carried out based on the eigenvalues of covariance matrix of received signal. The ratio of maximum-minimum eigenvalue of each sub-band is vulnerable to noise fluctuation. This paper suggests an optimized method to calculate the ratio of maximum-minimum eigenvalue. The simulation results imply that the design of digital channelizer can effectively separate the received signal with dynamically changeable sub-channel signals.     

Group behavior time series anomaly detection in specific network space based on separation degree

Specific network space, including virtual space and practical space, is a space for executing group behavior on specified regions via network. Due to the variability and unpredictability of time series in group behavior in special network space, the detection of normal and abnormal borders faces significant challenges. The parameters in traditional time series mode need to be predefined such as clustering method and anomaly detection methods science the results influentially depend on the selection of parameters. According to the characteristics of data, this paper proposes an efficient method called separation degree algorithm that can construct the self-adaptive interval based on the separation degree model to filter out anomaly data in virtual and practical spaces. The advantage allows us to automatically find the self-adaptive interval to improve the accuracy and applicability of anomaly detection based on the characteristics of the data instead of set parameters of traditional methods in network space. The extensive experimental result shows that the proposed method can effectively detect anomaly data from different spaces.     

Optimized Tunable Q Wavelet Transform Based Drowsiness Detection from Electroencephalogram Signals

Early discernment of drivers drowsy state may prevent numerous worldwide road accidents. Electroencephalogram (EEG) signals provide valuable information about the neurological changes for discrimination of alert and drowsy state. A signal is decomposed into multi-components for the analysis of the physiological state. Tunable Q wavelet transform (TQWT) decomposes the signal into low-pass and high-pass sub-bands without a choice of wavelet. The information content captured by these sub-bands depends on the choice of decomposition parameters. Due to the non-stationary nature of EEG signals, the predefined decomposition parameters of TQWT lead to information loss and degrade system performance. Hence it is required to automate the decomposition parameters in accordance with the nature of signals. In this paper, an optimized tunable Q wavelet transform (O-TQWT) is proposed for the adaptive selection of decomposition parameters by using different optimization algorithms. Objective function as a mean square error (MSE) of decomposition is minimized by optimization algorithms. Optimum decomposition parameters are used to decompose the signals into sub-bands. Time-domain based features are excerpted from the sub-bands of O-TQWT. Highly discriminant features selected by using Kruskal Wallis test are used as an input to different classification techniques. Classification accuracy of 96.14% is achieved by least square support vector machine with radial basis function kernel which is better than the other existing methodologies using the same database.     

Investigation of a two-point maximum entropy regularization method for signal enhancement applied to magnetoencephalography data

A two-point maximum entropy method (TPMEM) was investigated for post-acquisition signal recovery in magnetoencephalography (MEG) data, as a potential replacement of a low-pass (LP) filtering technique currently in use. We first applied TPMEM and the LP filter for signal recovery of synthetically noise corrupted MEG “phantom” data sets in which the true underlying signal was known. Results were quantified with the use of visual plots, percent error histograms, and the statistical parameters root mean squared error and Pearson’s correlation coefficient. Synthetically noise corrupted data from a simulated magnetic dipole was used to quantify the improvements gained in using TPMEM over LP filters in reconstructing known dipole parameters such as position, orientation, and magnitude. Finally, we applied TPMEM and LP filters to a sample MEG patient data set. Our results show that TPMEM has improved noise-reduction and signal recovery capabilities than those of the LP filter, and furthermore data processed with TPMEM shows less error in the reconstructed dipole parameters. We propose that TPMEM can be used for MEG signal processing, resulting in improved MEG source characterization.     

Computationally Efficient Cosine Modulated Filter Bank Design for ECG Signal Compression

In this work, computationally efficient and reliable cosine modulated filter banks (CMFBs) are designed for Electrocardiogram (ECG) data compression. First of all, CMFBs (uniform and non-uniform) are designed using interpolated finite impulse response (IFIR) prototype filter to reduce the computational complexity. To reduce the reconstruction error, linear iteration technique is applied to optimize the prototype filter. Then after, non-uniform CMFB is used for ECG data compression by decomposing ECG signal into various frequency bands. Subsequently, thresholding is applied for truncating the insignificant coefficients. The estimation of the threshold value is done by examining the significant energy of each band. Further, Run-length encoding (RLE) is utilized for improving the compression performance. The method is applied to MIT-BIH arrhythmia database for performance analysis of the proposed work. The experimental observations demonstrate that the proposed method has accomplished high compression ratio with the admirable quality of signal reconstruction. The proposed work provides the average values of compression ratio (CR), percent root mean square difference (PRD), percent root mean square difference normalized (PRDN), quality score (QS), correlation coefficient (CC), maximum error (ME), mean square error (MSE), and signal to noise ratio (SNR) are 23.86, 1.405, 2.55, 19.08, 0.999, 0.12, 0.054 and 37.611 dB, respectively. The proposed 8-channel uniform filter bank is used to detect the R-peak locations of the ECG signal. The comparative analysis shows that beats (locations and amplitudes) of both signals (original and reconstructed signals) are same.     

On-line optimization of biotechnological processes: I. Application to open algal pond

A new on-line optimization and control procedure applicable to biotechnological systems for which a precise mathematical model is unavailable has been developed and tested. The proposed approach is based on an online search for optimum operating conditions by an automatic system using a modified simplex algorithm to which several features have been added to permit real time operation. The simplex algorithm is the upper level of a hierarchical software package in which the other levels are cost evaluation, control, data acquisition, and signal processing. The optimization method was tested in a laboratory minipond for the cultivation of Spirulina platensis. The controlled parameters were light intensity, optical density, pH, and temperature. The proposed optimization method can be applied to other biological processes provided that the pertinent variables can be measured and controlled and the cost function can be defined mathematically.     

A Neural Network-Based Optimal Spatial Filter Design Method for Motor Imagery Classification

In this study, a novel spatial filter design method is introduced. Spatial filtering is an important processing step for feature extraction in motor imagery-based brain-computer interfaces. This paper introduces a new motor imagery signal classification method combined with spatial filter optimization. We simultaneously train the spatial filter and the classifier using a neural network approach. The proposed spatial filter network (SFN) is composed of two layers: a spatial filtering layer and a classifier layer. These two layers are linked to each other with non-linear mapping functions. The proposed method addresses two shortcomings of the common spatial patterns (CSP) algorithm. First, CSP aims to maximize the between-classes variance while ignoring the minimization of within-classes variances. Consequently, the features obtained using the CSP method may have large within-classes variances. Second, the maximizing optimization function of CSP increases the classification accuracy indirectly because an independent classifier is used after the CSP method. With SFN, we aimed to maximize the between-classes variance while minimizing within-classes variances and simultaneously optimizing the spatial filter and the classifier. To classify motor imagery EEG signals, we modified the well-known feed-forward structure and derived forward and backward equations that correspond to the proposed structure. We tested our algorithm on simple toy data. Then, we compared the SFN with conventional CSP and its multi-class version, called one-versus-rest CSP, on two data sets from BCI competition III. The evaluation results demonstrate that SFN is a good alternative for classifying motor imagery EEG signals with increased classification accuracy.     

Frequency limitations of the two-point central difference differentiation algorithm

A two-point central difference algorithm is often used to calculate the derivative of a function. This estimate is only valid over a limited frequency range. Therefore, the algorithm can be modeled as an ideal differentiator in series with a low-pass filter. The filter cutoff frequency is a function of the time between the points. We discuss the accuracy and limitations of using this algorithm on human saccadic eye movement data. To calculate the velocity of saccadic eye movements the algorithm should have a cutoff frequency of 74 Hz or above.     

Feature detection in human vision: a phase-dependent energy model

This paper presents a simple and biologically plausible model of how mammalian visual systems could detect and identify features in an image. We suggest that the points in a waveform that have unique perceptual significance as 'lines' and 'edges' are the points where the Fourier components of the waveform come into phase with each other. At these points 'local energy' is maximal. Local energy is defined as the square root of the sum of the squared response of sets of matched filters, of identical amplitude spectrum but differing in phase spectrum by 90 degrees: one filter type has an even-symmetric line-spread function, the other an odd-symmetric line-spread function. For a line the main contribution to the local energy peak is in the output of the even-symmetric filters, whereas for edges it is in the output of the odd-symmetric filters. If both filter types respond at the peak of local energy, both edges and lines are seen, either simultaneously or alternating in time. The model was tested with a series of images, and shown to predict well the position of perceived features and the organization of the images.     

The energetics and scaling of search strategies in bacteria

The influence of body size on the energetic cost of movement is well studied in animals but has been rarely investigated in bacteria. Here, I calculate the cost of four chemotactic strategies for different-sized bacteria by adding the costs of their locomotion and reorientation. Size differences of 0.1 microm result in 100,000-fold changes in the energetic cost of chemotaxis. The exact cost for any given size is a nonlinear function of flagella length, the minimum speed necessary to detect and respond to a signal, and the gradient of the signal. These parameters are interlinked in such a way that body size and strategy are tightly coupled to particular environmental gradients, offering avenues for explaining and exploring diversity and competition. The analysis here has implications beyond bacteria. Power-law regression through the minimum costs of transport for different kinds of chemotaxis has the same slope as that for swimming animals, suggesting a universal allometric equation for all swimming organisms.     

Detecting protein complexes from active protein interaction networks constructed with dynamic gene expression profiles

Background

Protein interaction networks (PINs) are known to be useful to detect protein complexes. However, most available PINs are static, which cannot reflect the dynamic changes in real networks. At present, some researchers have tried to construct dynamic networks by incorporating time-course (dynamic) gene expression data with PINs. However, the inevitable background noise exists in the gene expression array, which could degrade the quality of dynamic networkds. Therefore, it is needed to filter out contaminated gene expression data before further data integration and analysis.

Results

Firstly, we adopt a dynamic model-based method to filter noisy data from dynamic expression profiles. Then a new method is proposed for identifying active proteins from dynamic gene expression profiles. An active protein at a time point is defined as the protein the expression level of whose corresponding gene at that time point is higher than a threshold determined by a standard variance involved threshold function. Furthermore, a noise-filtered active protein interaction network (NF-APIN) is constructed. To demonstrate the efficiency of our method, we detect protein complexes from the NF-APIN, compared with those from other dynamic PINs.

Conclusion

A dynamic model based method can effectively filter out noises in dynamic gene expression data. Our method to compute a threshold for determining the active time points of noise-filtered genes can make the dynamic construction more accuracy and provide a high quality framework for network analysis, such as protein complex prediction.