Correlation Metrics Research Articles

Multi-level stacking of LSTM recurrent models for predicting stock-market indices

<abstract> <p>The ability to predict stock-market indices is important to investors and financial decision-makers. However, the uncertainty of available information makes accurate prediction extremely challenging. In this work, we propose and validate a multi-level stacking model of long short-term memory (LSTM) units for the short-term prediction of stock-index closing prices. The proposed machine-learning model is trained using historical data to predict next-day closing prices. The first layer of the multi-level stacked structure contains an ensemble of recurrent LSTM models that receives time-series data of historic opening, closing, high and low prices for current and previous days and outputs predictions about the next day's closing prices. The second and third layers consist of stacked multi-layer perceptron meta-models. We validated the new model on two stock indices, demonstrating its advantages over single-LSTM models. We also compared its performance against several extant statistical and machine-learning models on a subset of Standard & Poor's 500 index data between 2000 and 2016 using correlation and statistical metrics.</p> </abstract>

Prognostics With Variational Autoencoder by Generative Adversarial Learning

Prognostics predicts the future performance progression and remaining useful life (RUL) of in-service systems based on historical and contemporary data. One of the challenges in prognostics is the development of methods that are capable of handling real-world uncertainties that typically lead to inaccurate predictions. To alleviate the impacts of uncertainties and to achieve accurate degradation trajectory and RUL predictions, a novel sequence-to-sequence predictive model is proposed based on a variational autoencoder that is trained with generative adversarial networks. A long short-term memory network and a Gaussian mixture model are utilized as building blocks so that the model is capable of providing probabilistic predictions. Correlative and monotonic metrics are applied to identify sensitive features in the degradation progress, in order to reduce the uncertainty induced from raw data. Then, the selected features are concatenated with one-hot health state indicators as training data for the model to learn end of life without the need for prior knowledge of failure thresholds. Performance of the proposed model is validated by health monitoring data collected from real-world aeroengines, wind turbines, and lithium-ion batteries. The results demonstrate that significant performance improvement can be achieved in long-term degradation progress and RUL prediction tasks.

Towards a software defect proneness model: feature selection

This article is focused on improving static models of software reliability based on using machine learning methods to select the software code metrics that most strongly affect its reliability.The study used a merged dataset from the PROMISE Software Engineering repository, which contained data on testing software modules of fiveprograms and twenty-onecode metrics. For the prepared sampling, the most important features that affect the quality of software code have been selected using the following methods of feature selection: Boruta, Stepwiseselection, Exhaustive Feature Selection, Random Forest Importance, LightGBM Importance, Genetic Algorithms, Principal Component Analysis, Xverse python.Basing on the voting on the results of the work of the methods of feature selection, a static (deterministic) model of software reliability has been built, which establishes the relationship between the probability of a defect in the software module and the metrics of its code. It has been shown that this model includes such code metrics as branch countof a program, McCabe’s lines of codeand cyclomatic complexity, Halstead’s total number of operators and operands, intelligence, volume, andeffort value.A comparison of the effectiveness of different methods of feature selection has been put into practice, in particular, a study of the effectof the method of feature selection on the accuracy of classification using the following classifiers: Random Forest, Support Vector Machine, k-Nearest Neighbors, Decision Tree classifier, AdaBoost classifier, Gradient Boosting for classification. It has been shown that the use of any method of feature selection increases the accuracy of classification by at least ten percentcompared to the original dataset, which confirms the importance of this procedure for predicting software defects based on metric datasets that contain a significant number of highly correlated software code metrics.It has been found that the best accuracy of the forecast for most classifiers was reachedusing a set of features obtained from the proposed static model of software reliability. In addition, it has been shown that it is also possible to use separate methods, such as Autoencoder,Exhaustive Feature Selection and Principal Component Analysis withan insignificant loss of classification and prediction accuracy.

Developing test color samples to compute color fidelity of light sources for printing matter

Based on the clustering optimization of test color samples and a psychophysical experiment, the objective and subjective color fidelity of light sources for printing matter is evaluated, and an improved evaluation method thus has been proposed. Firstly, for representing the output characteristics of printing press, the International Color Consortium standard color target samples (ICC SCTS), which is measured in the process of color management, is used as a large color sample set. In a 6D spectral color space proposed in this paper, the optimized color sample set (OCSS) is obtained by Self Organizing Maps Neural Networks (SOMNN) clustering algorithm from ICC SCTS. Taking OCSS, ICC SCTS, and standard color sample set (SCSS) as the test color sample set, three objective color fidelity indexes (CFIs) CIE-Ra, Ra,2012, and CIE-Rf of 1202 light sources are calculated. The correlation metrics of the CFIs show that the OCSS highly improved the objective accuracy of color fidelity evaluation for printing matters. Secondly, in the psychophysical experiment, 20 observers have evaluated the visual color difference of the OCSS under the illumination of nine pairs of test and reference light sources. The subjective CFIs are calculated by using the visual color difference of OCSS obtained from the psychophysical experiment. In order to improve the subjective and objective consistency of CFIs, a polynomial modified model for objective color difference of OCSS is proposed. By the optimization of test color samples and the modification of color difference calculation, the method developed in this paper can be effectively and conveniently applied to the subjective and objective evaluation of light source for printing matters.

Improved fuzzy miner algorithm for business process discovery

Return material authorization (RMA) is a process in which a company decides to repair or replace customer’s defect product during the warranty period. To execute RMA, both company and customer obliged to follow standard operating procedure (SOP) which usually consists of many business processes of a company well. As the business process could cause inefficiencies, a company should improve their business process regularly. The best way is using process discovery. This research proposes a new improved fuzzy miner algorithm to represent binary correlation between activities. This new algorithm utilizes binary significance and binary correlation equally to acquire fuzzy model. While the original fuzzy miner algorithm uses various binary correlation metrics, the improved fuzzy miner algorithm uses only one metric and could capture the fuzzy model, accurately based on the event logs to capture more accurate business process model. In this research, ProM fuzzy miner is used as a comparison to the proposed improved time-based fuzzy miner. The results showed that the improved algorithm has higher value on conformance checking and able to capture business process model based on time interval, by using only time-interval significance as a binary correlation metrics.

Deep learning-based bone suppression in chest radiographs using CT-derived features: a feasibility study.

Bone suppression of chest X-ray holds the potential to improve the accuracy of target localization in image-guided radiation therapy (IGRT). However, the training dataset for bone suppression is limited because of the scarcity of bone-free radiographs. This study aims to develop a deep learning-based bone suppression method using CT-derived features to reduce the reliance on the bone-free dataset. In this study, 59 high-resolution lung CT scans were processed to generate the lung digital radiographs (DRs), bone DRs, and bone-free DRs, for the training and internal validation of the proposed cascade convolutional neural network (CCNN). A three-stage image processing framework (CT segmentation, DR simulation, and feature expansion) was developed to expand simulated lung DRs with different weightings of bone intensity. The CCNN consists of a bone detection network and a bone suppression network. In external validation, the trained CCNN was evaluated using 30 chest radiographs. The synthesized bone-suppressed radiographs were compared with the bone-suppressed reference in terms of peak signal-to-noise ratio (PSNR), mean absolute error (MAE), structural similarity index measure (SSIM), and Spearman's correlation coefficient. Furthermore, the effectiveness of the proposed feature expansion method and CCNN model were assessed via the ablation experiment and replacement experiment, respectively. Evaluation on real chest radiographs showed that the bone-suppressed chest radiographs closely matched with the bone-suppressed reference, achieving an accuracy of MAE =0.0087±0.0030, SSIM =0.8458±0.0317, correlation of 0.9554±0.0170, and PNSR of 20.86±1.60. After removing the feature expansion from the CCNN model, the performance decreased in terms of MAE (0.0294±0.0093, -237.9%), SSIM (0.7747±0.0.0416, -8.4%), correlation (0.8772±0.0271, -8.2%), and PSNR (15.53±1.42, -25.5%) metrics. We successfully demonstrated a novel deep learning-based bone suppression method using CT-derived features to reduce the reliance on the bone-free dataset. Implementation of the feature expansion procedures resulted in a remarkable reinforcement of the model performance. For the application of target localization in IGRT, the clinical testing of the proposed method in the context of radiation therapy is a necessary procedure to move from theory into practice.

Spatial point pattern analysis of traffic violations in Luzhou City, China

ABSTRACT The study presents spatial analysis of traffic violations in Luzhou city, China. Three metrics of spatial point pattern, namely intensity, spatial correlation, and spacing, are evaluated using (i) kernel density and quadrant count; (ii) K and L functions; and (iii) nearest neighbor and empty-space distance methods, respectively. The results show that: (i) three specific places have a high rate of illegal turning movements, disobeying prohibited signs, and illegal parking violations; (ii) all the violation categories have hot spots, among which 53.49% of the total appear at nine different places; (iii) the spatial correlation discloses the existence of spatial dependence and closeness of similar violations; and (iv) the spacing between traffic violations also signifies clustering of violations. Generally, the spatial patterns of traffic violations are nonuniform, have identifiable hot spots, and are clustered. The findings are beneficial for conceiving safety treatment strategies against traffic violations and additional relevant clustering studies.

Enhancing wall-to-wall forest structure mapping through detailed co-registration of airborne and terrestrial laser scanning data in Mediterranean forests

This paper presents a new co-registration procedure of complementary point clouds captured by both Terrestrial (TLS) and Airborne Laser Scanning (ALS) technologies. Starting from the geographic position of the TLS point cloud, a geometric features recognition algorithm, which evaluates digital terrain models obtained from both ALS and TLS, was developed and implemented in a new GIS software (ForeSight®). As a case study, we tested this new approach using point clouds acquired from both hand-held mobile TLS and ALS sensors over 24 test sites located in a protected area in southern Italy, with the ultimate goal of characterizing the different forest stand structures. From each aligned point cloud, a plot-level spatially explicit index (Enhanced Structural Spatial Index, ESCI) was derived to assess the three-dimensional structure of the considered forest stands. Then, we compared structural features derived from the ESCI index with different computed ALS metrics. Finally, the most correlated ALS metrics were used as predictors to produce an ESCI-map of the entire region of interest.

Challenges for variational reduced-density-matrix theory with three-particle N-representability conditions.

The direct variational optimization of the two-electron reduced density matrix (2RDM) can provide a reference-independent description of the electronic structure of many-electron systems that naturally capture strong or nondynamic correlation effects. Such variational 2RDM approaches can often provide a highly accurate description of strong electron correlation, provided that the 2RDMs satisfy at least partial three-particle N-representability conditions (e.g., the T2 condition). However, recent benchmark calculations on hydrogen clusters [N. H. Stair and F. A. Evangelista, J. Chem. Phys. 153, 104108 (2020)] suggest that even the T2 condition leads to unacceptably inaccurate results in the case of two- and three-dimensional clusters. We demonstrate that these failures persist under the application of full three-particle N-representability conditions (3POS). A variety of correlation metrics are explored in order to identify regimes under which 3POS calculations become unreliable, and we find that the relative squared magnitudes of the cumulant three- and two-particle reduced density matrices correlate reasonably well with the energy error in these systems. However, calculations on other molecular systems reveal that this metric is not a universal indicator for the reliability of the reduced-density-matrix theory with 3POS conditions.

On the Quality of Deep Representations for Kepler Light Curves Using Variational Auto-Encoders

Light curve analysis usually involves extracting manually designed features associated with physical parameters and visual inspection. The large amount of data collected nowadays in astronomy by different surveys represents a major challenge of characterizing these signals. Therefore, finding good informative representation for them is a key non-trivial task. Some studies have tried unsupervised machine learning approaches to generate this representation without much effectiveness. In this article, we show that variational auto-encoders can learn these representations by taking the difference between successive timestamps as an additional input. We present two versions of such auto-encoders: Variational Recurrent Auto-Encoder plus time (VRAEt) and re-Scaling Variational Recurrent Auto Encoder plus time (S-VRAEt). The objective is to achieve the most likely low-dimensional representation of the time series that matched latent variables and, in order to reconstruct it, should compactly contain the pattern information. In addition, the S-VRAEt embeds the re-scaling preprocessing of the time series into the model in order to use the Flux standard deviation in the learning of the light curves structure. To assess our approach, we used the largest transit light curve dataset obtained during the 4 years of the Kepler mission and compared to similar techniques in signal processing and light curves. The results show that the proposed methods obtain improvements in terms of the quality of the deep representation of phase-folded transit light curves with respect to their deterministic counterparts. Specifically, they present a good balance between the reconstruction task and the smoothness of the curve, validated with the root mean squared error, mean absolute error, and auto-correlation metrics. Furthermore, there was a good disentanglement in the representation, as validated by the Pearson correlation and mutual information metrics. Finally, a useful representation to distinguish categories was validated with the F1 score in the task of classifying exoplanets. Moreover, the S-VRAEt model increases all the advantages of VRAEt, achieving a classification performance quite close to its maximum model capacity and generating light curves that are visually comparable to a Mandel–Agol fit. Thus, the proposed methods present a new way of analyzing and characterizing light curves.

Principal Component Analysis of 1D 1H Diffusion Edited NMR Spectra of Protein Therapeutics.

The one-dimensional (1D) diffusion edited proton NMR method, Protein Fingerprint by Lineshape Enhancement (PROFILE) has been demonstrated to be suitable for higher order structure (HOS) characterization of protein therapeutics including monoclonal antibodies. Recent reports in the literature have demonstrated its advantages for HOS characterization over traditional methods such as circular dichroism and Fourier-transform infrared spectroscopy. Previously, we have demonstrated that the PROFILE method is complementary with high resolution 2D methyl correlated NMR methods and how both may be deployed as a multi-modal platform to further the utility of NMR for HOS characterization. A major limitation of the PROFILE method remains its need for high signal to noise data due to its reliance on convolution difference processing and linear correlation metrics to assess spectral similarity. Here we present an alternative method for analyzing 1D diffusion edited spectra, which overcomes this limitation by using nonlinear iterative partial least squares (NIPALS) principal component analysis, and which we dub PROtein Fingerprint Observed Using NIPALS Decomposition (PROFOUND). We demonstrate that results from the PROFOUND method are robust with respect to instrument, operator and in the presence of high experimental noise and how it may be employed to provide quantitative assessment of spectral similarity.

Synthetic pulmonary perfusion images from 4DCT for functional avoidance using deep learning

Purpose. To develop and evaluate the performance of a deep learning model to generate synthetic pulmonary perfusion images from clinical 4DCT images for patients undergoing radiotherapy for lung cancer. Methods. A clinical data set of 58 pre- and post-radiotherapy 99mTc-labeled MAA-SPECT perfusion studies (32 patients) each with contemporaneous 4DCT studies was collected. Using the inhale and exhale phases of the 4DCT, a 3D-residual network was trained to create synthetic perfusion images utilizing the MAA-SPECT as ground truth. The training process was repeated for a 50-imaging study, five-fold validation with twenty model instances trained per fold. The highest performing model instance from each fold was selected for inference upon the eight-study test set. A manual lung segmentation was used to compute correlation metrics constrained to the voxels within the lungs. From the pre-treatment test cases (N = 5), 50th percentile contours of well-perfused lung were generated from both the clinical and synthetic perfusion images and the agreement was quantified. Results. Across the hold-out test set, our deep learning model predicted perfusion with a Spearman correlation coefficient of 0.70 (IQR: 0.61–0.76) and a Pearson correlation coefficient of 0.66 (IQR: 0.49–0.73). The agreement of the functional avoidance contour pairs was Dice of 0.803 (IQR: 0.750–0.810) and average surface distance of 5.92 mm (IQR: 5.68–7.55). Conclusion. We demonstrate that from 4DCT alone, a deep learning model can generate synthetic perfusion images with potential application in functional avoidance treatment planning.

Assessment of wind and solar PV local complementarity for the hybridization of the wind power plants installed in Portugal

To decarbonize electrical power systems, it is essential to incorporate a high share of variable renewable energy sources while minimizing their costs. An important step towards this goal includes exploring the potential for the so-called hybrid renewable power plants through the combination of (existing) wind and solar power parks. Although that is not the usual approach in their design, hybrid power plants should take advantage of existing synergies, as generation complementarity among the renewable generation technologies, and benefit from using existing infrastructures and the same grid connection point.This investigation assesses the potential of existing Portuguese wind parks for hybridization with solar power photovoltaic generation. Correlation and energy metrics for assessing the complementarity at hourly and daily periods are applied to quantify existing synergies.Results show a high potential for hybrid power plants: levels of complementarity between wind and solar resources are globally high thus allowing to increase the share of variable renewable energy sources with minimal energy curtailment, when compared with scenarios of overplanting additional wind power capacity. The highest and most consistent levels of complementarity are found in the interior regions of central and northern Portugal.The research shows that Portugal has privileged weather conditions that allow the concept of hybrid power plants to be largely explored and extended on a large and efficient scale. This outcome, which may be replicated for other countries, is an important contribution to increasing the penetration of renewable energy in electrical power systems while minimizing the overall costs of renewable energy.

Estimating the Severity of Visual Field Damage From Retinal Nerve Fiber Layer Thickness Measurements With Artificial Intelligence.

PurposeThe purpose of this study was to assess the accuracy of artificial neural networks (ANN) in estimating the severity of mean deviation (MD) from peripapillary retinal nerve fiber layer (RNFL) thickness measurements derived from optical coherence tomography (OCT).MethodsModels were trained using 1796 pairs of visual field and OCT measurements from 1796 eyes to estimate visual field MD from RNFL data. Multivariable linear regression, random forest regressor, support vector regressor, and 1D convolutional neural network (CNN) models with sectoral RNFL thickness measurements were examined. Three independent subsets consisting of 698, 256, and 691 pairs of visual field and OCT measurements were used to validate the models. Estimation errors were visualized to assess model performance subjectively. Mean absolute error (MAE), root mean square error (RMSE), median absolute error, Pearson correlation, and R-squared metrics were used to assess model performance objectively.ResultsThe MAE and RMSE of the ANN model based on the testing dataset were 4.0 dB (95% confidence interval = 3.8–4.2) and 5.2 dB (95% confidence interval = 5.1–5.4), respectively. The ranges of MAE and RMSE of the ANN model on independent datasets were 3.3–5.9 dB and 4.4–8.4 dB, respectively.ConclusionsThe proposed ANN model estimated MD from RNFL measurements better than multivariable linear regression model, random forest, support vector regressor, and 1-D CNN models. The model was generalizable to independent data from different centers and varying races.Translational RelevanceSuccessful development of ANN models may assist clinicians in assessing visual function in glaucoma based on objective OCT measures with less dependence on subjective visual field tests.

Sensitivity of amplitude and phase based MEG measures of interhemispheric connectivity during unilateral finger movements

Interactions between different brain regions can be revealed by dependencies between their neuronal oscillations. We examined the sensitivity of different oscillatory connectivity measures in revealing interhemispheric interactions between primary motor cortices (M1s) during unilateral finger movements. Based on frequency, amplitude, and phase of the oscillations, a number of metrics have been developed to measure connectivity between brain regions, and each metric has its own strengths, weaknesses, and pitfalls. Taking advantage of the well-known movement-related modulations of oscillatory amplitude in M1s, this study compared and contrasted a number of leading connectivity metrics during distinct phases of oscillatory power changes. Between M1s during unilateral movements, we found that phase-based metrics were effective at revealing connectivity during the beta (15–35 Hz) rebound period linked to movement termination, but not during the early period of beta desynchronization occurring during the movement itself. Amplitude correlation metrics revealed robust connectivity during both periods. Techniques for estimating the direction of connectivity had limited success. Granger Causality was not well suited to studying these connections because it was strongly confounded by differences in signal-to-noise ratio linked to modulation of beta amplitude occurring during the task. Phase slope index was suggestive but not conclusive of a unidirectional influence between motor cortices during the beta rebound. Our findings suggest that a combination of amplitude and phase-based metrics is likely required to fully characterize connectivity during task protocols that involve modulation of oscillatory power, and that amplitude-based metrics appear to be more sensitive despite the lack of directional information.

Assessment of Intravertebral Mechanical Strains and Cancellous Bone Texture Under Load Using a Clinically Available Digital Tomosynthesis Modality.

Vertebral fractures are the most common osteoporotic fractures, but clinical means for assessment of vertebral bone integrity are limited in accuracy, as they typically use surrogate measures that are indirectly related to mechanics. The objective of this study was to examine the extent to which intravertebral strain distributions and changes in cancellous bone texture generated by a load of physiological magnitude can be characterized using a clinically available imaging modality. We hypothesized that digital tomosynthesis-based digital volume correlation (DTS-DVC) and image texture-based metrics of cancellous bone microstructure can detect development of mechanical strains under load. Isolated cadaveric T11 vertebrae and L2-L4 vertebral segments were DTS imaged in a nonloaded state and under physiological load levels. Axial strain, maximum principal strain, maximum compressive and tensile principal strains, and von Mises equivalent strain were calculated using the DVC technique. The change in textural parameters (line fraction deviation, anisotropy, and fractal parameters) under load was calculated within the cancellous centrum. The effect of load on measured strains and texture variables was tested using mixed model analysis of variance, and relationships of strain and texture variables with donor age, bone density parameters, and bone size were examined using regression models. Magnitudes and heterogeneity of intravertebral strain measures correlated with applied loading and were significantly different from background noise. Image texture parameters were found to change with applied loading, but these changes were not observed in the second experiment testing L2-L4 segments. DTS-DVC-derived strains correlated with age more strongly than did bone mineral density (BMD) for T11.

Identification of interdependencies and prediction of fault propagation for cyber–physical systems

Interdependence is an intrinsic feature of cyber–physical systems. Cyber and physical components are tightly integrated with each other, and hence, a trivial impairment in a part of the system may affect several components, leading to a sequence of failures that collapses the entire system. In this paper, we seek to identify the interdependencies among the components of a cyber–physical system using correlation metrics as well as a heuristic causation analysis method. We also demonstrate applicability of neural networks for prediction of imminent failures given the current system state. The proposed prediction tool can help system operators to perform timely preventive actions and mitigate the consequences of accidental failures and malicious attacks. As a case study, we have analyzed two smart grid test cases based on IEEE power bus systems, namely, IEEE-14 and IEEE-57.

Automated classification of brain tumors by deep learning-based models on magnetic resonance images using a developed web-based interface

Objective: Primary central nervous system tumors (PCNSTs) compose nearly 3% of newly diagnosed cancers worldwide and are more common in men. The incidence of brain tumors and PCNSTs-related deaths are gradually increasing all over the world. Recently, many studies have focused on automated machine learning (AutoML) algorithms which are developed using deep learning algorithms on medical imaging applications. The main purposes of this study are -to demonstrate the use of artificial intelligence-based techniques to predict medical images of different brain tumors (glioma, meningioma, pituitary adenoma) to provide technical support to radiologists, and -to develop a user-friendly and free web-based software to classify brain tumors for making quick and accurate clinical decisions. 
 
 Materials and Methods: Open-sourced T1-weighted magnetic resonance brain tumor images were achieved from Nanfang Hospital, Guangzhou, China, and General Hospital, Tianjin Medical University, To construct the proposed system which web-based interface and the deep learning-based models, the Keras/Auto-Keras library, which is employed in Python's programming language, is used. Accuracy, sensitivity, specificity, G-mean, F-score, and Matthews correlation coefficient metrics were used for performance evaluations.
 
 Results: While 80% (2599 instances) of the dataset was used in the training phase, 20% (465 instances) was employed in the testing phase. All the performance metrics were higher than 98% for the classification of brain tumors on the training data set. Similarly, all the evaluation metrics were higher than 91% except for sensitivity and MCC for meningioma on the testing dataset.
 
 Conclusion: The results from the experiment reveal that the proposed software can be used to detect and diagnose three types of brain tumors. This developed web-based software can be accessed freely in both English and Turkish at http://biostatapps.inonu.edu.tr/BTSY/.

The electrocardiographic forward problem: A benchmark study

BackgroundElectrocardiographic forward problems are crucial components for noninvasive electrocardiographic imaging (ECGI) that compute torso potentials from cardiac source measurements. Forward problems have few sources of error as they are physically well posed and supported by mature numerical and computational techniques. However, the residual errors reported from experimental validation studies between forward computed and measured torso signals remain surprisingly high. ObjectiveTo test the hypothesis that incomplete cardiac source sampling, especially above the atrioventricular (AV) plane is a major contributor to forward solution errors. MethodsWe used a modified Langendorff preparation suspended in a human-shaped electrolytic torso-tank and a novel pericardiac-cage recording array to thoroughly sample the cardiac potentials. With this carefully controlled experimental preparation, we minimized possible sources of error, including geometric error and torso inhomogeneities. We progressively removed recorded signals from above the atrioventricular plane to determine how the forward-computed torso-tank potentials were affected by incomplete source sampling. ResultsWe studied 240 beats total recorded from three different activation sequence types (sinus, and posterior and anterior left-ventricular free-wall pacing) in each of two experiments. With complete sampling by the cage electrodes, all correlation metrics between computed and measured torso-tank potentials were above 0.93 (maximum 0.99). The mean root-mean-squared error across all beat types was also low, less than or equal to 0.10 mV. A precipitous drop in forward solution accuracy was observed when we included only cage measurements below the AV plane. ConclusionFirst, our forward computed potentials using complete cardiac source measurements set a benchmark for similar studies. Second, this study validates the importance of complete cardiac source sampling above the AV plane to produce accurate forward computed torso potentials. Testing ECGI systems and techniques with these more complete and highly accurate datasets will improve inverse techniques and noninvasive detection of cardiac electrical abnormalities.

Spiking Neural Network for Augmenting Electroencephalographic Data for Brain Computer Interfaces.

With the advent of advanced machine learning methods, the performance of brain–computer interfaces (BCIs) has improved unprecedentedly. However, electroencephalography (EEG), a commonly used brain imaging method for BCI, is characterized by a tedious experimental setup, frequent data loss due to artifacts, and is time consuming for bulk trial recordings to take advantage of the capabilities of deep learning classifiers. Some studies have tried to address this issue by generating artificial EEG signals. However, a few of these methods are limited in retaining the prominent features or biomarker of the signal. And, other deep learning-based generative methods require a huge number of samples for training, and a majority of these models can handle data augmentation of one category or class of data at any training session. Therefore, there exists a necessity for a generative model that can generate synthetic EEG samples with as few available trials as possible and generate multi-class while retaining the biomarker of the signal. Since EEG signal represents an accumulation of action potentials from neuronal populations beneath the scalp surface and as spiking neural network (SNN), a biologically closer artificial neural network, communicates via spiking behavior, we propose an SNN-based approach using surrogate-gradient descent learning to reconstruct and generate multi-class artificial EEG signals from just a few original samples. The network was employed for augmenting motor imagery (MI) and steady-state visually evoked potential (SSVEP) data. These artificial data are further validated through classification and correlation metrics to assess its resemblance with original data and in-turn enhanced the MI classification performance.