Background Noise Distribution Research Articles

Brain Image Segmentation for Ultrascale Neuron Reconstruction via an Adaptive Dual-Task Learning Network.

Accurate morphological reconstruction of neurons in whole brain images is critical for brain science research. However, due to the wide range of whole brain imaging, uneven staining, and optical system fluctuations, there are significant differences in image properties between different regions of the ultrascale brain image, such as dramatically varying voxel intensities and inhomogeneous distribution of background noise, posing an enormous challenge to neuron reconstruction from whole brain images. In this paper, we propose an adaptive dual-task learning network (ADTL-Net) to quickly and accurately extract neuronal structures from ultrascale brain images. Specifically, this framework includes an External Features Classifier (EFC) and a Parameter Adaptive Segmentation Decoder (PASD), which share the same Multi-Scale Feature Encoder (MSFE). MSFE introduces an attention module named Channel Space Fusion Module (CSFM) to extract structure and intensity distribution features of neurons at different scales for addressing the problem of anisotropy in 3D space. Then, EFC is designed to classify these feature maps based on external features, such as foreground intensity distributions and image smoothness, and select specific PASD parameters to decode them of different classes to obtain accurate segmentation results. PASD contains multiple sets of parameters trained by different representative complex signal-to-noise distribution image blocks to handle various images more robustly. Experimental results prove that compared with other advanced segmentation methods for neuron reconstruction, the proposed method achieves state-of-the-art results in the task of neuron reconstruction from ultrascale brain images, with an improvement of about 49% in speed and 12% in F1 score.

Abstract 567: cMonitor: Comprehensive and sensitive monitoring of diverse cancer treatment outcomes by exome-wide mutation analysis in plasma cfDNA

Abstract Monitoring cancer patients for the early detection of minimal residual disease (MRD), cancer recurrence, and cancer progression is essential for assessing treatment response and predicting early relapse during/after treatment. Plasma cell-free DNA (cfDNA) provides unique opportunities for cancer monitoring given its non-invasive and comprehensive sampling of heterogeneous tumor clones. However, the low tumor content in cfDNA poses a major challenge for detecting tumor signals. Previous methods mostly rely on deep sequencing of small panels to capture the tumor signal. These methods usually require the labor-intensive design of customized gene panel, yet fail to identify the evolving tumor, which is essential for detecting second primary diseases or emerging subclones. To address these limitations, here we present OncoMonitor, a cancer monitoring method that comprehensively analyzes tumor mutations in cfDNA whole-exome sequencing data. Taking advantage of the availability of mutation information across the whole exome, our method (1) integrates all clonal tumor mutations identified from pre-treatment cfDNA samples (or tumor samples) to compensate for low tumor fraction in cfDNA, (2) suppresses sequencing errors at read level with an accurate random forest classifier to model the observed features from cfDNA fragments and further enhance the tumor signal, (3) builds sample-specific background noise distributions to predict recurrence and MRD to avoid interference from inter-individual variations and inter-experimental biases, and (4) detects tumor changes, especially second primary diseases, by identifying newly emerging tumor mutations de novo. Combining these techniques, for the first time, we show that OncoMonitor can sensitively and specifically detect both cancer recurrence/MRD and secondary primary cancers from plasma samples with low tumor fraction. Using simulated cfDNA sequence data with artificial mutation spike-ins, our method can detect recurrence at 0.025% tumor fraction with > 95% sensitivity and 95% specificity, and the second primary disease at 0.1% tumor fraction with around 75% sensitivity and 100% specificity. In a cohort of 9 non-small-cell lung cancer patients, we show that OncoMonitor can provide comprehensive tumor changes for treatment response prediction and capture emerging tumor clones; this cannot be achieved by previous methods that are based only on mutations in the pre-treatment surgery samples. In summary, with broad genomic sequencing coverage and comprehensive mutation analysis, our method can identify patients suffering from MRD or cancer recurrence/progression, provide a thorough view of their tumor status, and enable early intervention and personalized treatment. Citation Format: Shuo Li, Weihua Zeng, Xiaohui Ni, Mary L. Stackpole, Yonggang Zhou, Zorawar Noor, Zuyang Yuan, Edward B. Garon, Steven M. Dubinett, Wenyuan Li, Xianghong Zhou. cMonitor: Comprehensive and sensitive monitoring of diverse cancer treatment outcomes by exome-wide mutation analysis in plasma cfDNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 567.

Weak Lensing Study in VOICE Survey II: Shear Bias Calibrations

The VST Optical Imaging of the CDFS and ES1 Fields (VOICE) Survey is proposed to obtain deep optical $ugri$ imaging of the CDFS and ES1 fields using the VLT Survey Telescope (VST). At present, the observations for the CDFS field have been completed, and comprise in total about 4.9 deg$^2$ down to $r_\mathrm{AB}$$\sim$26 mag. In the companion paper by Fu et al. (2018), we present the weak lensing shear measurements for $r$-band images with seeing $\le$ 0.9 arcsec. In this paper, we perform image simulations to calibrate possible biases of the measured shear signals. Statistically, the properties of the simulated point spread function (PSF) and galaxies show good agreements with those of observations. The multiplicative bias is calibrated to reach an accuracy of $\sim$3.0%. We study the bias sensitivities to the undetected faint galaxies and to the neighboring galaxies. We find that undetected galaxies contribute to the multiplicative bias at the level of $\sim$0.3%. Further analysis shows that galaxies with lower signal-to-noise ratio (SNR) are impacted more significantly because the undetected galaxies skew the background noise distribution. For the neighboring galaxies, we find that although most have been rejected in the shape measurement procedure, about one third of them still remain in the final shear sample. They show a larger ellipticity dispersion and contribute to $\sim$0.2% of the multiplicative bias. Such a bias can be removed by further eliminating these neighboring galaxies. But the effective number density of the galaxies can be reduced considerably. Therefore efficient methods should be developed for future weak lensing deep surveys.

Performance analysis of diffusion least mean fourth algorithm over network

The diffusion least mean fourth (DLMF) algorithm was initially presented for distributed estimate over network with the sub-Gaussian node noise, which can obtain better performance than diffusion least mean square (DLMS) algorithm. Unfortunately, due to the higher order moment of the weight error in the DLMF algorithm and complexity of the network, it is difficult to analyze the mean square performance of the DLMF algorithm. In this paper, the stochastic behaviors including mean behavior and mean square behavior are investigated when the input signals of all nodes are Gaussian processes and the background noises have symmetric probability density functions. The analytical models of the mean weight error and the mean square deviation (MSD) are derived by using a novel manner, which can accurately predict the algorithm behavior. The simulations are carried out to verify the validity of the analysis and compare the performance of DLMF algorithm with that of the DLMS algorithm under different background noise distributions.

Cepstrum-Domain Model Combination Based on Decomposition of Speech and Noise Using MMSE-LSA for ASR in Noisy Environments

This paper presents an efficient method for combining models of speech and noise for robust speech recognition applications in noisy environments. This method decomposes the cepstrum domain representation of noise-corrupted speech into clean speech cepstrum and background noise cepstrum components using a minimum mean squared error-log spectral amplitude (MMSE-LSA) criterion. Speech recognition is then performed on noisy cepstrum domain observations using a model that is formed by parallel combination of cepstrum domain clean speech distributions and background noise distributions estimated using this MMSE-LSA based noise decomposition. This method is far more efficient than other parallel model combination (PMC) procedures because model combination is performed directly in the cepstrum domain rather than in the linear spectral domain. Whereas background noise model estimation is addressed as a separate issue in existing PMC procedures, this method explicitly incorporates a mechanism to continually update background noise models and signal-to-noise ratio (SNR) estimates over time. The performance of the proposed cepstrum-domain model combination method is compared with a well known implementation of PMC which uses a log-normal approximation when combining speech and background noise model means and variances on a connected digit string recognition task which is subjected to mismatched channel and environment conditions. As a result, it is shown that the proposed model combination technique gives a word error rate that is comparable to PMC when background noise information and SNR are known prior to estimation. The paper will also present the results of experiments where a combination of cepstrum-domain feature compensation and model combination are applied to this task.

Influence of multichannel combination, parallel imaging and other reconstruction techniques on MRI noise characteristics

The statistical properties of background noise such as its standard deviation and mean value are frequently used to estimate the original noise level of the acquired data. This requires the knowledge of the statistical intensity distribution of the background signal, that is, the probability density of the occurrence of a certain signal intensity. The influence of many new MRI techniques and, in particular, of various parallel-imaging methods on the noise statistics has neither been rigorously investigated nor experimentally demonstrated yet. In this study, the statistical distribution of background noise was analyzed for MR acquisitions with a single-channel and a 32-channel coil, with sum-of-squares (SoS) and spatial-matched-filter (SMF) data combination, with and without parallel imaging using k-space and image-domain algorithms, with real-part and conventional magnitude reconstruction and with several reconstruction filters. Depending on the imaging technique, the background noise could be described by a Rayleigh distribution, a noncentral chi-distribution or the positive half of a Gaussian distribution. In particular, the noise characteristics of SoS-reconstructed multichannel acquisitions (with k-space-based parallel imaging or without parallel imaging) differ substantially from those with image-domain parallel imaging or SMF combination. These effects must be taken into account if mean values or standard deviations of background noise are employed for data analysis such as determination of local noise levels. Assuming a Rayleigh distribution as in conventional MR images or a noncentral chi-distribution for all multichannel acquisitions is invalid in general and may lead to erroneous estimates of the signal-to-noise ratio or the contrast-to-noise ratio.

Influence of high-resolution processing on background noise distribution in ship-borne radar

When a high-resolution algorithm is applied in a ship-borne radar, its non-linearity and distributional characteristics before high-resolution processing determine the distributional characteristics of background clutter after high resolution and detector design afterwards. Because background noise before high-resolution has physical significance, the statistical model of first-order Bragg lines and second-order components of sea clutter is put forward. Then, by using higher order cumulative quantity's statistical verification of actual measured data, it is concluded that the background noise before high resolution conforms to normal distribution in a ship-borne radar. The non-linearity of the high-resolution algorithm means that the background noise after high-resolution processing conforms to non-normal distribution. Non-normal-distributed clutter mainly includes Weibull, lognormal, and K-clutter. Rayleigh clutter can be seen as a special case of Weibull clutter. These clutter types have different statistical characteristics and can be discriminated by recognition of clutter characteristics. The numerical domain's distribution after high-resolution processing is determined by an improved, minimum entropy, clutter characteristics recognition method based on the AIC rule, namely, a two-parameter domain-scanning method. This identification method has a higher recognition rate. It is verified that the background noise after high-resolution by pre-whitened-constrained-MUSIC conforms to lognormal distribution.

Multivariate analysis of neuronal interactions in the generalized partial least squares framework: simulations and empirical studies

Identification of spatiotemporal interactions within/between neuron populations is critical for detection and characterization of large-scale neuronal interactions underlying perception, cognition, and behavior. Univariate analysis has been employed successfully in many neuroimaging studies. However, univariate analysis does not explicitly test for interactions between distributed areas of activity and is not sensitive to distributed responses across the brain. Multivariate analysis can explicitly test for multiple statistical models, including the designed paradigm, and allows for spatial and temporal model detection. Here, we investigate multivariate analysis approaches that take into consideration the 4D (time and space) covariance structure of the data. Principal component analysis (PCA) and independent component analysis (ICA) are two popular multivariate approaches with distinct mathematical constraints. Common difficulties in using these two different decompositions include the following: classification of the revealed components (task-related signal versus noise), overall signal-to-noise sensitivity, and the relatively low computational efficiency (multivariate analysis requires the entire raw data set and more time for model identification analysis). Using both Monte Carlo simulations and empirical data, we derived and tested the generalized partial least squares (gPLS) framework, which can incorporate both PCA and ICA decompositions with computational efficiency. The gPLS method explicitly incorporates the experimental design to simplify the identification of characteristic spatiotemporal patterns. We performed parametric modeling studies of a blocked-design experiment under various conditions, including background noise distribution, sampling rate, and hemodynamic response delay. We used a randomized grouping approach to manipulate the degrees of freedom of PCA and ICA in gPLS to characterize both paradigm coherent and transient brain responses. Simulation data suggest that in the gPLS framework, PCA mostly outperforms ICA as measured by the receiver operating curves (ROCs) in SNR from 0.01 to 100, the hemodynamic response delays from 0 to 3 TR in fMRI, background noise models of Guassian, sub-Gaussian, and super-Gaussian distributions and the number of observations from 5, 10, to 20 in each block of a six-block experiment. Further, due to selective averaging, the gPLS method performs robustly in low signal-to-noise ratio (<1) experiments. We also tested PCA and ICA using PLS in a simulated event-related fMRI data to show their similar detection. Finally, we tested our gPLS approach on empirical fMRI motor data. Using the randomized grouping method, we are able to identify both transient responses and consistent paradigm/model coherent components in the 10-epoch block design motor fMRI experiment. Overall, studies of synthetic and empirical data suggest that PLS analysis, using PCA decomposition, provides a stable and powerful tool for exploration of fMRI/behavior data.

Robust statistical methods for reducing sferics noise contaminating transient electromagnetic measurements

The transient electromagnetic (TEM) method is used extensively for mineral exploration and other applications such as geothermal soundings, oil exploration, groundwater pollution, soil salinity and geological mapping. Sferics pulses produced by lightning strokes propagating in the ionosphere‐earth waveguide cavity induce noise in a bandwidth of a few Hz to tens of kHz. The usual method of stacking and calculating the mean of a given stack cannot effectively reduce the spike‐like noise induced by high‐amplitude sferics pulses. To reduce this type of noise, a number of different ways of stacking data were investigated and compared. Noise data were stacked by using robust estimators such as the median, trimmed mean, and a range of M‐estimators. Since storage of all the samples of a given stack can take up a prohibitively large amount of microprocessor memory, recursive algorithms for the M‐estimators and their standard error were developed for the real‐time reduction of sferics pulses. The recursive algorithms have been demonstrated to work effectively on windowed data, and thus the memory normally required to obtain the mean is sufficient for calculation of the M‐estimate. In the recursive calculation of the robust estimate of the transient response, the spread of the background noise distribution (known as the scale of the data) needs to be known or calculated. In the algorithm that has been developed, the scale of the data is derived from a noise run carried out before pulsing the transmitter loop with current. It has been assumed that the presence of a signal does not change the scale of the data. This value of the scale of the data has been used to obtain a robust estimate of the transient response itself. To allow for possible changes in the background noise level during a given survey, the estimate of the scale of the data is updated throughout the survey. Many tests of the performance of the recursive algorithms have been carried out with both simulated noise data and sferics data that have been recorded previously on magnetic tape. The results show that for sferics activity as high as that observed in northern latitudes of Australia in summer, a noise reduction by a factor of about 5 (when compared with simple stacking) should be obtained. In areas where sferics noise predominates over geological background signal, such a reduction should lead to an increase in target detection depth by approximately 50%.

Moving ordered statistics CFAR detection for nonhomogeneous backgrounds

A new robust order statistic-based CFAR scheme, the moving ordered statistics (MOS) CFAR detector, is proposed. The MOSCFAR detector effectively estimates the noise level in nonhomogeneous backgrounds by using the spatial subspace concept and the order statistics estimation method. Like the OS-CFAR detector, the MOS-CFAR processor selects one of the reference cells as the estimate of the background noise power. However, the rank of the selected sample is a variable and varies depending on the different background noise distributions in the reference cells. The detection performances of the MOSCFAR processor is investigated and compared with other well known CFAR detectors by means of simulations. The results show that the proposed MOS-CFAR detector is superior to the other schemes in regulating the false alarm rate in nonhomogeneous backgrounds.

Efficient importance sampling techniques for simulation of multiuser communication systems

The problem of simulating error rates in direct sequence spread-spectrum code division multiple-access (CDMA) systems is examined. Due to the computational complexity required to simulate these systems, an importance sampling technique is developed based upon previous work by the authors. A conditional weighting function is derived such that the linear shift class of biasing densities can be employed. Results are given for a variety of detector structures and background noise distributions. It is shown that this biasing scheme can dramatically reduce the run time of realistic multiple-access simulations. >

Estimation of Area-Averaged Rainfall over Tropical Oceans from Microwave Radiometry: A Single Channel Approach

This paper presents a new simple retrieval algorithm for estimating area-time averaged rain rates over tropical oceans by using single channel microwave measurements from satellites. The algorithm was tested by using the Nimbus-5 Electrically Scanning Microwave Radiometer and a simple microwave radiative transfer model to retrieve seasonal 5-deg x 5-deg area averaged rainrate over the tropical Atlantic and Pacific from December 1973 to November 1974. The brightness temperatures were collected and analyzed into histograms for each season and in each grid box from December 1973 to November 1974. The histograms suggest a normal distribution of background noise plus a skewed rain distribution at the higher brightness temperatures. By using a statistical estimation procedure based upon normally distributed background noise, the rain distribution was separated from the raw histogram. The radiative transfer model was applied to the rain-only distribution to retrieve area-time averaged rainrates throughout the tropics. Despite limitations of single channel information, the retrieved seasonal rain rates agree well in the open ocean with expectations based upon previous estimates of tropical rainfall over the oceans.

Statistical evaluation of cell kinetic data from DNA flow cytometry (FCM) by the EM algorithm.

Flow cytometric DNA measurements yield the amount of DNA for each of a large number of cells. A DNA histogram normally consists of a mixture of one or more constellations of G0/G1-, S-, G2/M-phase cells, together with internal standards, debris, background noise, and one or more populations of clumped cells. We have modelled typical DNA histograms as a mixed distribution with Gaussian densities for the G0/G1 and G2/M phases, an S-phase density, assumed to be uniform between the G0/G1 and G2/M peaks, observed with a Gaussian error, and with Gaussian densities for standards of chicken and trout red blood cells. The debris is modelled as a truncated exponential distribution, and we also have included a uniform background noise distribution over the whole observation interval. We have explored a new approach for maximum-likelihood analyses of complex DNA histograms by the application of the EM algorithm. This algorithm was used for four observed DNA histograms of varying complexity. Our results show that the algorithm works very well, and it converges to reasonable values for all parameters. In simulations from the estimated models, we have investigated bias, variance, and correlations of the estimates.

Noise-Correlation Functions for Anisotropic Noise Fields

Under certain conditions, the signal-to-noise gain of a hydrophone array can be expressed in terms of signal- and noise-correlation functions. When ideal propagation is assumed, the signal correlation is well-defined and the problem is reduced to establishing the noise-correlation function. Noise-correlation functions for two-dimensional and three-dimensional isotropic-noise fields are well-known: namely, the functions J0(X) and sinX/X. In this paper, an equation for the noise-correlation function is derived that is valid for any orientation of receiver elements and any distribution of background noise. Assuming a background-noise field whose power in a given direction is proportional to the radius vector of an oblate spheroid, an assumption in qualitative agreement with recent theoretical and experimental results, the noise-correlation functions for horizontal elements and for vertical elements are explored in detail. Quantitative results are given for noise fields of different anisotropy. [This work was supported by the Bureau of Ships, U. S. Navy.]