Noise Correction Research Articles

Measuring decay time in non-diffuse rooms

Measurements in non-diffuse rooms present challenges in determining decay time. A room with a severe flutter echo may exhibit a double slope decay without the incorporation of a coupled chamber. The measured decay time may differ greatly depending on the software used to measure, the choice of Schroeder integral sampling range, and how noise correction is implemented. A standardized approach to characterizing decay time under non-diffuse conditions is needed. In this talk, the authors discuss possible methods for determining the optimal noise correction and sampling of the Schroeder integral, including varying the IR truncation and sampling start and stop range. Computational optimization methods are employed to objectively determine the most reasonable settings for interpreting a particular impulse response.

Correlation between subcutaneous adipose tissue of the head and body mass index: Implications for functional neuroimaging.

High body mass index (BMI) is generally assumed to represent overall amounts of body adipose tissue (fat). Increased adipose tissue amounts in persons with increased BMI has been cited as a barrier to assessment of body tissues such as muscle. Significant increases in the amount of adipose tissue between the dermal layer and the skull may result in high electrical impedance and/or increased light diffusion causing a lower signal to noise ratio during use of neuroimaging tools such as electroencepholography (EEG), transcranial direct current stimulation (tDCS), and functional near infrared spectroscopy (fNIRS). Investigating how subcutaneous adipose tissue in the head region increases with respect to total body fat percentage and BMI is an important step in developing mathematical corrections in neuroimaging measurements as BMI increases, as recommended in other measurement modalities such as electromyography (EMG). We hypothesized that percentage of subcutaneous adipose tissue in the head region would increase with respect to both total body fat percentage and BMI. A statistically significant increase in subcutaneous head fat percentage occurred with increased BMI and total body fat percentage. The data investigated in this study indicate that participant age, sex, and BMI are important features to consider in model corrections during data signal processing and analyses for subcutaneous head fat in neuroimaging approaches. The data in this project serve to provide physiological justification for this practice along with regression analyses to be considered for physiologically-based signal to noise correction algorithms.

Constant amplitude modulation heterodyne interferometry.

Laser interferometer with picometer precision is a key technology in the space gravitational wave detection. Many interferometry strategies have been put forward for the multiple purposes in the past 10 yr. We propose a new interferometry method, called constant amplitude modulation (CAM) heterodyne interferometry. Differently, the CAM provides an optical pilot tone (OPT) for the noise correction theme. Compared with the analog pilot tone, the OPT can record and correct more noises, such as the analog to digital converter sampling jitter, the photodetector noise and the analog front-end noise. From the discussion, the modulated depth ϕmodulate=1.375 rad and the power ratio of the beam split n=0.432 are the best choice for the CAM-modulated parameter. Moreover, a simulated case has been implemented for the verification of the CAM strategy. Therefore, the CAM gives us another excellent choice in the optical design of the interferometer.

Instance difficulty-based noise correction for crowdsourcing

Crowdsourcing offers an efficient way to obtain a multiple noisy label set of each instance from different crowd workers and then label integration algorithms are used to infer its integrated label. In spite of the effectiveness of label integration algorithms, there always exists a certain degree of noise in the integrated labels, and thus noise correction algorithms have been proposed to reduce the effect of noise. However, existing noise correction algorithms seldom consider the effect of instance difficulty on noise correction. In this paper, we argue that the greater the difficulty of an instance, the fewer crowd workers can label it correctly, and the more likely the instance is a noise instance. Based on this premise, we propose a simple but very effective noise correction algorithm called instance difficulty-based noise correction (IDNC). In IDNC, we at first propose two methods to measure the difficulty of each instance. Then, we use the proposed two methods to filter the noise instances to obtain a clean set and a noise set. Finally, we build two different classifiers on the clean set to correct the noise instances in the noise set via the consensus voting. The extensive experiments on both simulated and real-world crowdsourced datasets validated the effectiveness and efficiency of our proposed IDNC.

Label noise and self-learning label correction in cardiac abnormalities classification

Objective. Learning to classify cardiac abnormalities requires large and high-quality labeled datasets, which is a challenge in medical applications. Small datasets from various sources are often aggregated to meet this requirement, resulting in a final dataset prone to label noise due to inter- and intra-observer variability and different expertise. It is well known that label noise can affect the performance and generalizability of the trained models. In this work, we explore the impact of label noise and self-learning label correction on the classification of cardiac abnormalities on large heterogeneous datasets of electrocardiogram (ECG) signals. Approach. A state-of-the-art self-learning multi-class label correction method for image classification is adapted to learn a multi-label classifier for electrocardiogram signals. We evaluated our performance using 5-fold cross-validation on the publicly available PhysioNet/Computing in Cardiology (CinC) 2021 Challenge data, with full and reduced sets of leads. Due to the unknown label noise in the testing set, we tested our approach on the MNIST dataset. We investigated the performance under different levels of structured label noise for both datasets. Main results. Under high levels of noise, the cross-validation results of self-learning label correction show an improvement of approximately 3% in the challenge score for the PhysioNet/CinC 2021 Challenge dataset and an improvement in accuracy of 5% and reduction of the expected calibration error of 0.03 for the MNIST dataset. We demonstrate that self-learning label correction can be used to effectively deal with the presence of unknown label noise, also when using a reduced number of ECG leads.

Pushing the colorimetry camera-based fluorescence microscopy to low light imaging by denoising and dye combination.

Colorimetry camera-based fluorescence microscopy (CCFM) is a single-frame imaging method for observing multiple biological events simultaneously. Compared with the traditional multi-color fluorescence microscopy methods based on sequential excitation or spectral splitting, the CCFM method simplifies multi-color fluorescence imaging experiments, while keeping a high spatial resolution. However, when the level of the detected fluorescence signal decreases, the image quality, the demosaicking algorithm precision, and the discrimination of fluorescence channels on the colorimetry camera will also decrease. Thus, CCFM has a poor color resolution under a low signal level. For example, the crosstalk will be higher than 10% when the signal is less than 100 photons/pixel. To solve this problem, we developed a new algorithm that combines sCMOS noise correction with demosaicking, and a dye selection method based on the spectral response characteristics of the colorimetry camera. By combining the above two strategies, low crosstalk can be obtained with 4 ∼ 6 fold fewer fluorescence photons, and low light single-frame four-color fluorescence imaging was successfully performed on fixed cos-7 cells. This study expands the power of the CCFM method, and provides a simple and efficient way for various bioimaging applications in low-light conditions.

Vibration Compensation for a Vehicle-Mounted Atom Gravimeter

The performance of absolute atom gravimeters used on moving platforms, such as vehicles, ships and aircrafts, is strongly affected by the vibration noise. To suppress its influence, we introduce a vibration compensation method, in which a classical accelerometer is used to measure the vibration noise. The measurement results obtained by the accelerometer show the vibration noise in the vehicle can be 2 orders of magnitude greater than that in the lab during daytime, and can induce an interferometric phase fluctuation with a standard deviation of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$16.70\pi $ </tex-math></inline-formula> . With the compensation method, our vehicle-mounted atom gravimeter can work normally under harsh conditions. Comparing the Allan standard deviations before and after the vibration noise correction, we find a suppression factor of 22.74 can be achieved in static condition with a half interrogation time of T = 20 ms, resulting in a sensitivity of 1.35 mGal/Hz <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1/2</sup> , and a standard deviation of 0.5 mGal with an average time of 10 s. We also demonstrate the first test of an atom gravimeter in a moving vehicle, in which a suppression factor of 50.85 and a sensitivity of 60.88 mGal/Hz <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1/2</sup> were realized with T = 5 ms.

Noise-Robust Learning from Multiple Unsupervised Sources of Inferred Labels

Deep Neural Networks (DNNs) generally require large-scale datasets for training. Since manually obtaining clean labels for large datasets is extremely expensive, unsupervised models based on domain-specific heuristics can be used to efficiently infer the labels for such datasets. However, the labels from such inferred sources are typically noisy, which could easily mislead and lessen the generalizability of DNNs. Most approaches proposed in the literature to address this problem assume the label noise depends only on the true class of an instance (i.e., class-conditional noise). However, this assumption is not realistic for the inferred labels as they are typically inferred based on the features of the instances. The few recent attempts to model such instance-dependent (i.e., feature-dependent) noise require auxiliary information about the label noise (e.g., noise rates or clean samples). This work proposes a theoretically motivated framework to correct label noise in the presence of multiple labels inferred from unsupervised models. The framework consists of two modules: (1) MULTI-IDNC, a novel approach to correct label noise that is instance-dependent yet not class-conditional; (2) MULTI-CCNC, which extends an existing class-conditional noise-robust approach to yield improved class-conditional noise correction using multiple noisy label sources. We conduct experiments using nine real-world datasets for three different classification tasks (images, text and graph nodes). Our results show that our approach achieves notable improvements (e.g., 6.4% in accuracy) against state-of-the-art baselines while dealing with both instance-dependent and class-conditional noise in inferred label sources.

Noise correction of large deviations with anomalous scaling.

We present a path integral calculation of the probability distribution associated with the time-integrated moments of the Ornstein-Uhlenbeck process that includes the Gaussian prefactor in addition to the dominant path or instanton term obtained in the low-noise limit. The instanton term was obtained recently [D. Nickelsen and H. Touchette, Phys. Rev. Lett. 121, 090602 (2018)0031-900710.1103/PhysRevLett.121.090602] and shows that the large deviations of the time-integrated moments are anomalous in the sense that the logarithm of their distribution scales nonlinearly with the integration time. The Gaussian prefactor gives a correction to the low-noise approximation and leads us to define an instanton variance giving some insights as to how anomalous large deviations are created in time. The results are compared with simulations based on importance sampling, extending our previous results based on direct Monte Carlo simulations. We conclude by explaining why many of the standard analytical and numerical methods of large deviation theory fail in the case of anomalous large deviations.

Unbiased line edge roughness measurement using profile-averaging method for precise roughness parameters measurement

Background: Scanning electron microscopy (SEM)-based line edge roughness (LER) measurement suffers from an error due to noise in SEM image. Noise correction methods have been developed to obtain unbiased roughness results, however, there are still concerns in the viewpoint of measurement precision. Aim: To develop an unbiased roughness analysis for highly precise LER measurement. Approach: Combining the conventional unbiased roughness analysis with a profile-averaging method, where a line pattern is repeatedly measured and then the obtained profiles are aligned and averaged. The experimental result measured by SEM was verified using atomic force microscopy (AFM)-based LER metrology, which has higher reliability than SEM. Results: The experimental evaluation showed that the proposed method can obtain roughness parameters more precisely than the conventional method. Conclusions: When the noise in the line edge profile by SEM is too large, it is necessary to reduce the noise beforehand and then perform roughness analysis in order to obtain precise roughness results. The proposed method enables to measure LER with the highest precision using SEM. Additionally, the AFM-based LER metrology was demonstrated as a feasible technique to evaluate the performance of SEM-based LER metrology.

The Optimal Determination of the Truncation Time of Non-Exponential Sound Decays

The noise effects in the room impulse response (RIR) make the decay range of the integrated impulse response insufficient for reliable determination of reverberation time (RT). One of the preferred techniques to minimize noise effects is based on noise subtraction, RIR truncation, and correction for the truncation. The success of RT estimation through the method depends critically on the accurate estimation of the truncation time (TT). However, noise fluctuation and RIR irregularities can lead to discrepancies in the determined TT from the optimal value. The general goal of this paper is to improve RT estimates. An iterative procedure based on a non-exponential decay model consisting of a double-slope decay term and a noise term is presented to estimate the TT accurately. The model parameters are generated until the iterative procedure converges to a minimum difference between the energy decay curve (EDC) generated by the model and the Schroeder decay function. The decay rates of the EDCs with added pink noise levels are compared to those of the EDCs with low background noise. In addition, the detected TTs and the corresponding RTs are compared with the existing method and the noise compensation method (subtraction–truncation–correction method).

Feasibility of Deep Learning-Based Noise and Artifact Reduction in Coronal Reformation of Contrast-Enhanced Chest Computed Tomography.

This study aimed to evaluate the feasibility of a deep learning method for imaging artifact and noise reduction in coronal reformation of contrast-enhanced chest computed tomography (CT). A total of 19,052 coronal reformatted chest CT images of 110 CT image sets (55 pairs of concordant 16- and 320-row CT image sets) were included and used to train a deep learning algorithm for artifact and noise correction. For internal validation, 4093 coronal reformatted CT images of 25 patients from 16-row CT images underwent correction processing. For external validation, chest CT images of 30 patients (1028 coronal reformatted CT images), acquired in other institutions using different scanners, were subjected to correction processing. For both validations, image quality was compared between original ("CT origin ") and deep learning-based corrected ("CT correct ") CT images. Quantitative analysis for stair-step artifact (coefficient of variance of CT density on coronal reformation), image noise, signal-to-noise ratio, and contrast-to-noise ratio were evaluated. Subjective image quality scores were assigned for image contrast, artifact, and conspicuity of major structures. CT correct showed significantly reduced stair-step artifact (mean coefficient of variance: CT origin 7.35 ± 2.0 vs CT correct 5.17 ± 2.4, P < 0.001) and image noise and improved signal-to-noise ratio and contrast-to-noise ratio in the aorta, pulmonary artery, and liver, compared with those of CT origin ( P < 0.01). On subjective analysis, CT correct had higher image contrast, lower artifact, and better conspicuity than CT origin . Most results of the external validation were consistent with those obtained from the internal validation, except for those concerning the pulmonary artery. Deep learning-based artifact correction significantly improved the image quality of coronal reformation chest CT by reducing image noise and artifacts.

Exciting a cavitating tip vortex with synthetic inflow turbulence: A CFD analysis of vortex kinematics, dynamics and sound generation

Cavitating tip vortices are one of the main contributors to underwater radiated noise (URN). To predict URN and evaluate propeller designs, it is necessary to predict cavity dynamics. To this end, a tip vortex generated by an elliptical wing is simulated in wetted and cavitating conditions, using scale-resolving simulations. The vortex is excited by synthetic inflow turbulence with varying inflow turbulence intensities. Vortex kinematics and cavity dynamics are analysed, and validated against experiments and a semi-analytical model from literature. The far-field radiated noise is analysed using an acoustic analogy. Using a background noise correction, the sound due to inflow turbulence is removed, and the sound due to cavity dynamics is isolated. Based on the sound spectra, the main noise generating mechanisms are identified. Cavitating simulations predict an increase in far-field radiated noise of approximately 15 dB, while doubling the inflow turbulence intensity results in an increase of approximately 10 dB.

KnotAli: informed energy minimization through the use of evolutionary information

BackgroundImproving the prediction of structures, especially those containing pseudoknots (structures with crossing base pairs) is an ongoing challenge. Homology-based methods utilize structural similarities within a family to predict the structure. However, their prediction is limited to the consensus structure, and by the quality of the alignment. Minimum free energy (MFE) based methods, on the other hand, do not rely on familial information and can predict structures of novel RNA molecules. Their prediction normally suffers from inaccuracies due to their underlying energy parameters.ResultsWe present a new method for prediction of RNA pseudoknotted secondary structures that combines the strengths of MFE prediction and alignment-based methods. KnotAli takes a multiple RNA sequence alignment as input and uses covariation and thermodynamic energy minimization to predict possibly pseudoknotted secondary structures for each individual sequence in the alignment. We compared KnotAli’s performance to that of three other alignment-based programs, two that can handle pseudoknotted structures and one control, on a large data set of 3034 RNA sequences with varying lengths and levels of sequence conservation from 10 families with pseudoknotted and pseudoknot-free reference structures. We produced sequence alignments for each family using two well-known sequence aligners (MUSCLE and MAFFT).ConclusionsWe found KnotAli’s performance to be superior in 6 of the 10 families for MUSCLE and 7 of the 10 for MAFFT. While both KnotAli and Cacofold use background noise correction strategies, we found KnotAli’s predictions to be less dependent on the alignment quality. KnotAli can be found online at the Zenodo image: https://doi.org/10.5281/zenodo.5794719

DeU-Net 2.0: Enhanced deformable U-Net for 3D cardiac cine MRI segmentation.

Automatic segmentation of cardiac magnetic resonance imaging (MRI) facilitates efficient and accurate volume measurement in clinical applications. However, due to anisotropic resolution, ambiguous borders and complicated shapes, existing methods suffer from the degradation of accuracy and robustness in cardiac MRI segmentation. In this paper, we propose an enhanced Deformable U-Net (DeU-Net) for 3D cardiac cine MRI segmentation, composed of three modules, namely Temporal Deformable Aggregation Module (TDAM), Enhanced Deformable Attention Network (EDAN), and Probabilistic Noise Correction Module (PNCM). TDAM first takes consecutive cardiac MR slices (including a target slice and its neighboring reference slices) as input, and extracts spatio-temporal information by an offset prediction network to generate fused features of the target slice. Then the fused features are also fed into EDAN that exploits several flexible deformable convolutional layers and generates clear borders of every segmentation map. A Multi-Scale Attention Module (MSAM) in EDAN is proposed to capture long range dependencies between features of different scales. Meanwhile, PNCM treats the fused features as a distribution to quantify uncertainty. Experimental results show that our DeU-Net achieves the state-of-the-art performance in terms of the commonly used evaluation metrics on the Extended ACDC dataset and competitive performance on other two datasets, validating the robustness and generalization of DeU-Net.

DRUMMER-rapid detection of RNA modifications through comparative nanopore sequencing.

MotivationThe chemical modification of ribonucleotides regulates the structure, stability and interactions of RNAs. Profiling of these modifications using short-read (Illumina) sequencing techniques provides high sensitivity but low-to-medium resolution i.e. modifications cannot be assigned to specific transcript isoforms in regions of sequence overlap. An alternative strategy uses current fluctuations in nanopore-based long read direct RNA sequencing (DRS) to infer the location and identity of nucleotides that differ between two experimental conditions. While highly sensitive, these signal-level analyses require high-quality transcriptome annotations and thus are best suited to the study of model organisms. By contrast, the detection of RNA modifications in microbial organisms which typically have no or low-quality annotations requires an alternative strategy. Here, we demonstrate that signal fluctuations directly influence error rates during base-calling and thus provides an alternative approach for identifying modified nucleotides.ResultsDRUMMER (Detection of Ribonucleic acid Modifications Manifested in Error Rates) (i) utilizes a range of statistical tests and background noise correction to identify modified nucleotides with high confidence, (ii) operates with similar sensitivity to signal-level analysis approaches and (iii) correlates very well with orthogonal approaches. Using well-characterized DRS datasets supported by independent meRIP-Seq and miCLIP-Seq datasets we demonstrate that DRUMMER operates with high sensitivity and specificity.Availability and implementationDRUMMER is written in Python 3 and is available as open source in the GitHub repository: https://github.com/DepledgeLab/DRUMMER.Supplementary information Supplementary data are available at Bioinformatics online.

Variance reducing and noise correction in protein quantification by measuring fluctuations in fluorescence due to photobleaching

Quantifying the absolute protein number using the ratio between the variance and the mean of the protein Fluorescence intensity is a straightforward method for microscopy imaging. Recently, this method has been expanded to fluorescence decaying processes due to photobleaching with binomial distribution. The article examines the method proposed and shows how it can be adapted to the case of variance in the initial number of proteins between the cells. The article shows that the method can be improved by the implementation of the information processing of each frame independently from other frames. By doing so, the variance in determining the protein number can be reduced. In addition, the article examines the management of unwanted noises in the measurement, offers a solution for the shot noise and background noise, examines the expected error caused by the decay constant inaccuracy, and analyzes the expected difficulties in conducting a practical experiment, which includes a non-exponential decay and variance in the photobleaching rate of the cells. The method can be applied to any superposition of n 0 discrete decaying processes. However, the evaluation of expected errors in quantification is essential for early planning of the experimental conditions and evaluation of the error.

Novel hybrid ensemble credit scoring model with stacking-based noise detection and weight assignment

Credit scoring is used to help financial institutions control default risks and reduce economic losses, and a variety of mainstream machine learning and data mining algorithms have been applied for this purpose. However, real-world datasets are generally noisy, which seriously affects the performance of credit scoring models. Among the mainstream strategies for handling noise, instance filtering may result in information loss, especially for hard-to-access credit datasets, and label noise correction may produce erroneous information in the dataset. In this study, to reduce the adverse impact of noisy data on the performance of classification algorithms, a novel hybrid ensemble credit scoring model with stacking-based noise detection and weight assignment is developed to remove or adapt noisy data in raw datasets and to form noise-detected training data to obtain excellent default risk prediction competence. Furthermore, a new weight assignment approach based on a cloud model is proposed, which is applied to calculate the weight values of the classifiers in the weighted voting ensemble model to improve the prediction accuracy of the proposed model. In this study, five public datasets are adopted using five performance metrics to evaluate the performance of the proposed model. The experimental results demonstrate good model prediction power and robustness.

Fast, Accurate, and Robust T2 Mapping of Articular Cartilage by Neural Networks.

For T2 mapping, the underlying mono-exponential signal decay is traditionally quantified by non-linear Least-Squares Estimation (LSE) curve fitting, which is prone to outliers and computationally expensive. This study aimed to validate a fully connected neural network (NN) to estimate T2 relaxation times and to assess its performance versus LSE fitting methods. To this end, the NN was trained and tested in silico on a synthetic dataset of 75 million signal decays. Its quantification error was comparatively evaluated against three LSE methods, i.e., traditional methods without any modification, with an offset, and one with noise correction. Following in-situ acquisition of T2 maps in seven human cadaveric knee joint specimens at high and low signal-to-noise ratios, the NN and LSE methods were used to estimate the T2 relaxation times of the manually segmented patellofemoral cartilage. In-silico modeling at low signal-to-noise ratio indicated significantly lower quantification error for the NN (by medians of 6–33%) than for the LSE methods (p < 0.001). These results were confirmed by the in-situ measurements (medians of 10–35%). T2 quantification by the NN took only 4 s, which was faster than the LSE methods (28–43 s). In conclusion, NNs provide fast, accurate, and robust quantification of T2 relaxation times.

Methods and algorithms for protecting information in optical text recognition systems

The subject of the study. A concept of OCR systems performance improvement is proposed, which is achieved through the integrated use of special algorithms for preliminary processing of documents picture, an extended set of service functions, and advanced techniques for information protection. Study objectives: development of algorithms that compensate for the influence of the unfavorable points like imperfect lighting conditions overshooting, images geometric deformation, noises, etc., which corrupt the pictured text, on the efficiency of that text recognition. It is needed to provide for a series of service procedures that would ensure adequate data handling while viewing, converting, and storing in standard formats the results, and ensuring the possibility to exchange data in open communication networks. Additionally, it is necessary to ensure the information protection against unauthorized use at the stage of data processing and provide secretiveness of their transmission through the communication channels. Investigation methods and research results: developed and tested algorithms for preliminary picture data processing, namely, for the captured image geometry transformation, picture noise correction with different filters, image binarization when using the adaptive thresholds reduced the negative influence of irregular image portions illumination; in the software, the special services ensure the data processing ease and information protection are affected. In particular, the interactive procedure for text segmentation is proposed, which implies the possibility of anonymizing its fragments and contributes to collecting confidentiality for documents treated. The package for processing document shots contains the face detection algorithm bringing the identification of such information features; it can be used further in the task of face recognition. After the textual doc is recognized, the received data encryption is provided by generating a QR-code and the steganography methods can deliver the privacy of this information transmission. The algorithms' structures are described in detail and the stability of their work under various conditions is investigated. Focused on the case study, docs' text recognition software was developed with the usage of Tesseract version 4.0 optical character recognition program. The program named "HQ Scanner" is written in Python using the present resources of the OpenCV library. An original technique for evaluating the efficiency of algorithms using the criterion of the maximum probability of correct text recognition is implemented in the software. Conclusions. The study results can constitute the basis for developing advanced specialized software for easy-to-use commercial OCR systems.