Abstract Beta-band burst activity is a key biomarker of Parkinsonian pathophysiology and a control signal for adaptive deep brain stimulation (aDBS). Existing burst detection approaches rely on bandpass filtering followed by envelope extraction and thresholding, which introduces onset/offset latency due to sliding-window estimation and can be unstable under variations in signal-to-noise ratio or threshold choice. We introduce a switching state space approach for real-time burst detection that models oscillatory activity as a superposition of latent harmonic components and explicitly represents burst and non-burst regimes. The framework performs real-time inference using a set of Kalman filters and computes posterior mode probabilities using a Markov transition prior, producing sample-by-sample burst probabilities together with phase and uncertainty estimates. In simulated data with sinusoidal beta bursts embedded in both white and pink noise, the method improved burst detection accuracy, reduced onset latency and offset latency compared with causal envelope-based baselines, and showed reduced sensitivity to threshold selection. In analysis of real sub-thalamic nucleus recordings during a grip-force task, the MSSR method recovered movement-related burst modulation consistent with prior reports. Our results indicate that state space switching provides a principled route to low-latency burst detection that may better support closed-loop stimulation strategies timed to beta bursts.

Data-driven signal decomposition methods decompose a signal into its underlying components in a flexible and adaptive way, taking into account the signal characteristics. Here we focus on univariate signals, whose decomposition is ill-posed. Classical univariate approaches – like variational mode decomposition – constrain the solution space (e.g. through narrowband priors), and often require the number of components to be known in advance. These assumption, however, limit the algorithm's usage in certain real-life applications. Instead, we exploit the flexibility of neural networks to replace fixed (narrowband) priors with data-driven priors. Our model, called Iterative Deep-Learning-Based Signal Decomposition (IDSD), iteratively extracts an adaptive number of various types of components from a signal, with no restrictions on the bandwidth of a component. We show superior performance of IDSD both in a controlled setup with synthetic data, and on two real datasets concerning tidal waves and physiological measurements.

For patients experiencing myocardial infarction (MI), localizing the affected cardiac region using electrocardiography (ECG) can reduce the time to reperfusion therapy, reducing morbidity and mortality. Extracting relevant information from ECG signals is not trivial, and computational methods have been developed aiming to assist physicians in making faster and better decisions in emergency situations. However, their clinical adoption remains limited due to the high false alarm rates consequence of the low generalizability of these methods. This research compares the performance of three machine learning techniques — Lasso, Support Vector Machine, and Gradient Boosting Machine — with varying degrees of complexity in localizing MI. Vectorcardiography-derived features were used as input to the models due to their ability to capture spatial and temporal information regarding the heart’s electrical activity. An autoencoder was employed to smooth the feature space, facilitating more efficient model training and improving generalization. To further address generalizability challenges, an inter-patient validation approach was employed. Models were trained on the PTB-XL dataset and externally validated on the PTB Diagnostic dataset. Results demonstrate that Lasso, a simpler model, achieved the highest AUC of 0.74 on the external dataset, outperforming more complex models such as SVM (0.72) and GBM (0.68). Also, the combination of Lasso with the autoencoder provided superior generalization compared to other state-of-the-art methods reported in the MI localization literature. This highlights the proposed method’s suitability for clinical settings, where model generalizability and reliability are critical. Furthermore, our method offers the advantage of explainability, allowing the extraction of clinical and physiological insights from the data and bridging the gap between computational methods and clinicians.

Complexity and signal recurrence metrics obtained from body surface potential mapping (BSPM) allow quantifying atrial fibrillation (AF) substrate complexity. This study aims to correlate electrocardiographic imaging (ECGI) detected reentrant patterns with BSPM-calculated signal complexity and recurrence metrics. BSPM signals were recorded from 28 AF patients (17 male, 11 women, 62.69 ± 8.09 y.o.), followed by ECGI calculation. Signal complexity and recurrence metrics were computed on BSPM and ECGI signals. Rotors per second and rotor duration were computed on ECGI signals for each atrium and the whole atrial surface. Correlation between BSPM metrics and ECGI reentrant patterns for the entire atrial surface and for left atrium (LA) and right atrium (RA) were analyzed. Atrial complexity and recurrence metrics strongly correlated when computed on BSPM and ECGI. Higher sample entropy and relative harmonic energy (RHE) correlated with rotors of short duration. The highest dominant frequency of the ECGI signals did not correlate with the reentrant activity of the ECGI. Higher short- and long-term recurrence of BSPM signals correlated with longer duration rotors, particularly for long-term recurrence (rLA=0.74 vs. rRA=0.42). Only ECGI-based reentrant parameters showed higher LA complexity compared to RA (p < 0.05). BSPM metrics strongly correlate with metrics measured on ECGI signals. BSPM metrics indicate a more elevated atrial electro-structural remodeling aligned with more short-duration rotors from ECGI computations. Although BSPM delivers qualitative AF reentry data, ECGI remains essential for identifying regional substrate complexity.

Electrocardiographic imaging (ECGI) aims to noninvasively estimate heart surface potentials starting from body surface potentials. This is classically based on geometric information on the torso and the heart from imaging, which complicates clinical application. In this study, we aim to develop a deep learning framework to estimate heart surface potentials solely from body surface potentials, enabling wider clinical use. The framework introduces two main components: the transformation of 3D torso and heart geometries into standard 2D representations, and the development of a customized deep learning network model. The 2D torso and heart representations maintain a consistent layout across different subjects, making the proposed framework applicable to different torso-heart geometries. With spatial information incorporated in the 2D representations, the torso-heart physiological relationship can be learnt by the network. The deep learning model is based on a Pix2Pix network, adapted to work with 2.5D data in our task, i.e., 2D body surface potential maps (BSPMs) and 2D heart surface potential maps (HSPMs) with time sequential information. We propose a new loss function tailored to this specific task, which uses a cosine similarity and different weights for different inputs. BSPMs and HSPMs from 11 healthy subjects (8 females and 3 males) and 29 idiopathic ventricular fibrillation (IVF) patients (11 females and 18 males) were used in this study. Performance was assessed on a test set by measuring the similarity and error between the output of the proposed model and the solution provided by mainstream ECGI, by comparing HSPMs, the concatenated electrograms (EGMs), and the estimated activation time (AT) and recovery time (RT). The mean of the mean absolute error (MAE) for the HSPMs was 0.012 ± 0.011, and the mean of the corresponding structural similarity index measure (SSIM) was 0.984 ± 0.026. The mean of the MAE for the EGMs was 0.004 ± 0.004, and the mean of the corresponding Pearson correlation coefficient (PCC) was 0.643 ± 0.352. Results suggest that the model is able to precisely capture the structural and temporal characteristics of the HSPMs. The mean of the absolute time differences between estimated and reference activation times was 6.048 ± 5.188 ms, and the mean of the absolute differences for recovery times was 18.768 ± 17.299 ms. Overall, results show similar performance between the proposed model and standard ECGI, exhibiting low error and consistent clinical patterns, without the need for CT/MRI. The model shows to be effective across diverse torso-heart geometries, and it successfully integrates temporal information in the input. This in turn suggests the possible use of this model in cost effective clinical scenarios like patient screening or post-operative follow-up.

The visualization and comparison of electrophysiological information in the atrium among different patients could be facilitated by a standardized 2D atrial mapping. However, due to the complexity of the atrial anatomy, unfolding the 3D geometry into a 2D atrial mapping is challenging. In this study, we aim to develop a standardized approach to achieve a 2D atrial mapping that connects the left and right atria, while maintaining fixed positions and sizes of atrial segments across individuals. Atrial segmentation is a prerequisite for the process. Segmentation includes 19 different segments with 12 segments from the left atrium, 5 segments from the right atrium, and two segments for the atrial septum. To ensure consistent and physiologically meaningful segment connections, an automated procedure is applied to open up the atrial surfaces and project the 3D information into 2D. The corresponding 2D atrial mapping can then be utilized to visualize different electrophysiological information of a patient, such as activation time patterns or phase maps. This can in turn provide useful information for guiding catheter ablation. The proposed standardized 2D maps can also be used to compare more easily structural information like fibrosis distribution with rotor presence and location. We show several examples of visualization of different electrophysiological properties for both healthy subjects and patients affected by atrial fibrillation. These examples show that the proposed maps provide an easy way to visualize and interpret intra-subject information and perform inter-subject comparison, which may provide a reference framework for the analysis of the atrial fibrillation substrate before treatment, and during a catheter ablation procedure.

AIMS:Although electrical activity of the normal human heart is well characterized by the electrocardiogram, detailed insights into within-subject and between-subject variations of ventricular activation and recovery by noninvasive electroanatomic mapping are lacking. We characterized human epicardial activation and recovery within and between normal subjects using non-invasive electrocardiographic imaging (ECGI) as a basis to better understand pathology. METHODS AND RESULTS:Epicardial activation and recovery were assessed by ECGI in 22 normal subjects, 4 subjects with bundle branch block (BBB) and 4 with long-QT syndrome (LQTS). We compared characteristics between the ventricles [left ventricle (LV) and right ventricle (RV)], sexes, and age groups (<50/≥50years). Pearson's correlation coefficient (CC) was used for within-subject and between-subject comparisons. Age of normal subjects averaged 49 ± 14 years, 6/22 were male, and no structural/electrical heart disease was present. The average activation time was longer in LV than in RV, but not different by sex or age. Electrical recovery was similar for the ventricles, but started earlier and was on average shorter in males. Median CCs of between-subject comparisons of the ECG signals, activation, and recovery patterns were 0.61, 0.32, and 0.19, respectively. Within-subject beat-to-beat comparisons yielded higher CCs (0.98, 0.89, and 0.82, respectively). Activation and/or recovery patterns of patients with BBB or LQTS contrasted significantly with those found in the normal population. CONCLUSION:Activation and recovery patterns vary profoundly between normal subjects, but are stable individually beat to beat, with a male preponderance to shorter recovery. Individual characterization by ECGI at baseline serves as reference to better understand the emergence, progression, and treatment of electrical heart disease.

Tensor-based signal decomposition methods offer a promising avenue for signal decomposition of short, non-stationary and non-linear input signals. A novel Tensor-based Singular Spectrum Decomposition (TSSD) framework is presented that extends Singular Spectrum Decomposition (SSD) to tensors for univariate signals using two tensor decomposition techniques, namely, Multilinear Singular Value Decomposition (MLSVD) and Canonical Polyadic Decomposition (CPD). Results indicate improved performance under the influence of noise and in the presence of sizeable trends. Experiments on real-life data on EEG signals from epileptic seizures further show the strong practical relevance of TSSD as a tool for exploratory signal analysis that helps unveil underlying system(s) in signals.

Objective: A growing body of research focuses on the automated diagnosis of acute myocardial infarction (AMI) using electrocardiogram (ECG) recordings. Several methods rely on differences between the ECG at baseline (no AMI) and during AMI condition. However, this approach may not sufficiently account for the progress of AMI, and it can underestimate the effect of false positives in a continuous monitoring setting. This in turn may hinder the adoption of automated methods for AMI diagnosis in the clinical practice. In this study, we propose a new automated method for the dynamic assessment of AMI condition. This method accounts for the dynamic nature underlying AMI events and the need for a low false positives incidence. Using a reduced 3-lead ECG system, we developed a novel set of parameters able to capture changes over time in the distribution properties of ECG -derived features. These parameters are used to train and validate a deep learning model in order to perform dynamic assessment of AMI condition. Conclusion: Results suggest that the proposed method is able to capture the dynamic evolution of AMI with a false positive rate below 1%. Significance: Thanks to the reduced number of leads, the proposed method could be used to assess AMI condition in long-term, remote and home monitoring, and intensive care unit (ICU) environments.

The electrocardiogram (ECG) is the standard method in clinical practice to non-invasively analyze the electrical activity of the heart, from electrodes placed on the body’s surface. The ECG can provide a cardiologist with relevant information to assess the condition of the heart and the possible presence of cardiac pathology. Nonetheless, the global view of the heart’s electrical activity given by the ECG cannot provide fully detailed and localized information about abnormal electrical propagation patterns and corresponding substrates on the surface of the heart. Electrocardiographic imaging, also known as the inverse problem in electrocardiography, tries to overcome these limitations by non-invasively reconstructing the heart surface potentials, starting from the corresponding body surface potentials, and the geometry of the torso and the heart. This problem is ill-posed, and regularization techniques are needed to achieve a stable and accurate solution. The standard approach is to use zero-order Tikhonov regularization and the L-curve approach to choose the optimal value for the regularization parameter. However, different methods have been proposed for computing the optimal value of the regularization parameter. Moreover, regardless of the estimation method used, this may still lead to over-regularization or under-regularization. In order to gain a better understanding of the effects of the choice of regularization parameter value, in this study, we first focused on the regularization parameter itself, and investigated its influence on the accuracy of the reconstruction of heart surface potentials, by assessing the reconstruction accuracy with high-precision simultaneous heart and torso recordings from four dogs. For this, we analyzed a sufficiently large range of parameter values. Secondly, we evaluated the performance of five different methods for the estimation of the regularization parameter, also in view of the results of the first analysis. Thirdly, we investigated the effect of using a fixed value of the regularization parameter across all reconstructed beats. Accuracy was measured in terms of the quality of reconstruction of the heart surface potentials and estimation of the activation and recovery times, when compared with ground truth recordings from the experimental dog data. Results show that values of the regularization parameter in the range (0.01–0.03) provide the best accuracy, and that the three best-performing estimation methods (L-Curve, Zero-Crossing, and CRESO) give values in this range. Moreover, a fixed value of the regularization parameter could achieve very similar performance to the beat-specific parameter values calculated by the different estimation methods. These findings are relevant as they suggest that regularization parameter estimation methods may provide the accurate reconstruction of heart surface potentials only for specific ranges of regularization parameter values, and that using a fixed value of the regularization parameter may represent a valid alternative, especially when computational efficiency or consistency across time is required.

The liver is the primary site for the metabolism and detoxification of many compounds, including pharmaceuticals. Consequently, it is also the primary location for many adverse reactions. As the liver is not readily accessible for sampling in humans; rodent or cell line models are often used to evaluate potential toxic effects of a novel compound or candidate drug. However, relating the results of animal and in vitro studies to relevant clinical outcomes for the human in vivo situation still proves challenging. In this study, we incorporate principles of transfer learning within a deep artificial neural network allowing us to leverage the relative abundance of rat in vitro and in vivo exposure data from the Open TG-GATEs data set to train a model to predict the expected pattern of human in vivo gene expression following an exposure given measured human in vitro gene expression. We show that domain adaptation has been successfully achieved, with the rat and human in vitro data no longer being separable in the common latent space generated by the network. The network produces physiologically plausible predictions of human in vivo gene expression pattern following an exposure to a previously unseen compound. Moreover, we show the integration of the human in vitro data in the training of the domain adaptation network significantly improves the temporal accuracy of the predicted rat in vivo gene expression pattern following an exposure to a previously unseen compound. In this way, we demonstrate the improvements in prediction accuracy that can be achieved by combining data from distinct domains.

Background: The optimal way to determine repolarization time (RT) from the intracardiac unipolar electrogram (UEG) has been a topic of debate for decades. RT is typically determined by either the Wyatt method or the “alternative method,” which both consider UEG T-wave slope, but differently.Objective: To determine the optimal method to measure RT on the UEG.Methods: Seven pig hearts surrounded by an epicardial sock with 100 electrodes were Langendorff-perfused with selective cannulation of the left anterior descending (LAD) coronary artery and submersed in a torso-shaped tank containing 256 electrodes on the torso surface. Repolarization was prolonged in the non-LAD-regions by infusing dofetilide and shortened in the LAD-region using pinacidil. RT was determined by the Wyatt (tWyatt) and alternative (tAlt) methods, in both invasive (recorded with epicardial electrodes) and in non-invasive UEGs (reconstructed with electrocardiographic imaging). tWyatt and tAlt were compared to local effective refractory period (ERP).Results: With contact mapping, mean absolute error (MAE) of tWyatt and tAlt vs. ERP were 21 ms and 71 ms, respectively. Positive T-waves typically had an earlier ERP than negative T-waves, in line with theory. tWyatt -but not tAlt-shortened by local infusion of pinacidil. Similar results were found for the non-invasive UEGs (MAE of tWyatt and tAlt vs. ERP were 30 ms and 92 ms, respectively).Conclusion: The Wyatt method is the most accurate to determine RT from (non) invasive UEGs, based on novel and historical analyses. Using it to determine RT could unify and facilitate repolarization assessment and amplify its role in cardiac electrophysiology.

Drug-induced intrahepatic cholestasis (DIC) is a main type of hepatic toxicity that is challenging to predict in early drug development stages. Preclinical animal studies often fail to detect DIC in humans. In vitro toxicogenomics assays using human liver cells have become a practical approach to predict human-relevant DIC. The present study was set up to identify transcriptomic signatures of DIC by applying machine learning algorithms to the Open TG-GATEs database. A total of nine DIC compounds and nine non-DIC compounds were selected, and supervised classification algorithms were applied to develop prediction models using differentially expressed features. Feature selection techniques identified 13 genes that achieved optimal prediction performance using logistic regression combined with a sequential backward selection method. The internal validation of the best-performing model showed accuracy of 0.958, sensitivity of 0.941, specificity of 0.978, and F1-score of 0.956. Applying the model to an external validation set resulted in an average prediction accuracy of 0.71. The identified genes were mechanistically linked to the adverse outcome pathway network of DIC, providing insights into cellular and molecular processes during response to chemical toxicity. Our findings provide valuable insights into toxicological responses and enhance the predictive accuracy of DIC prediction, thereby advancing the application of transcriptome profiling in designing new approach methodologies for hazard identification.

Previous studies have shown that global short- (ST) and long-term (LT) atrial fibrillation (AF) dynamics can be characterized non-invasively by Body Surface Potentials (BSPM). Also, Electrocardiographic Imaging (ECGI) may add information as it can characterize locally the atrial substrate. The objective of this study is to compare AF dynamics characterized on both BSPM and ECGI signals. Two consecutive 4-second BSPM signals from 34 AF patients (23 male, 8 paroxysmal, 63.1 ± 9.5 years) were recorded, followed by ECGI computation. ST and LT dynamics metrics were computed in both BSPM and ECGI, assessed from a multivariate autocorrelation of the signals. BSPM features of ST dynamics positively correlated with LT dynamics (0.52 and 0.78). Analogous values of correlation were obtained in ECGI. When normalized by the LT dynamics, the ST inversely correlated with the speed of propagation of AF at half AF cycle (BSPM, r = −0.28 vs. ECGI, r = −0.45), showing higher stability in the ST propagation for faster AF. BSPM and ECGI reflected similar relationships in the analysis of AF propagation dynamics. Results were consistent with previous studies and suggest that BSPM are sufficient to characterize global AF dynamics, while ECGI may become relevant when more localized information is required.

The theory of orthogonal multiwavelets offers enhanced flexibility for signal processing applications and analysis by employing multiple waveforms simultaneously, rather than a single one. When implementing them with polyphase filter banks, it has been recognized that balanced vanishing moments are needed to prevent undesirable artifacts to occur, which otherwise compromise the interpretation and usefulness of the multiwavelet analysis. In the literature, several such balanced orthogonal multiwavelets have been constructed and published; but however useful, their choice is still limited. In this work we present a full parameterization of the space of all orthogonal multiwavelets with two balanced vanishing moments (of orders 0 and 1), for arbitrary given multiplicity and degree of the polyphase filter. This allows one to search for matching multiwavelets for a given application, by optimizing a suitable design criterion. We present such a criterion, which is sparsity-based and useful for detection purposes, which we illustrate with an example from electrocardiographic signal analysis. We also present explicit conditions to build in a third balanced vanishing moment (of order 2), which can be used as a constraint together with the earlier parameterization. This is demonstrated by constructing a balanced orthogonal multiwavelet of multiplicity three, having three balanced vanishing moments, but this approach can easily be employed for arbitrary multiplicity.

Microvascular changes may represent an underlying mechanism through which exposure to fine particulate matter with a diameter ≤ 2.5 µm (PM2.5) contributes to age-related disease development. We investigated the effect of recent and chronic exposure to PM2.5 on the microcirculation, exemplified by retinal vessel diameters, using repeated measurements in 8- to 12-year-old children.221 children (49.1% girls; mean age 9.9 years) were examined repeatedly (25 one, 124 two, and 72 three times) adding up to 489 retinal vessel examinations. Same-day exposure to PM2.5 was measured at school. In addition, recent (same and previous day) and chronic (yearly mean) exposure was modelled at the child's residence using a high-resolution interpolation model. Residential proximity to major roads was also assessed. Changes in retinal vessel diameters associated with recent and chronic exposures were estimated using mixed models, while adjusting for other known covariates such as sex, age, BMI, blood pressure and birth weight.Each 10 µg/m³ increment in same-day exposure to PM2.5 measured at school was associated with 0.35 µm (95% CI: 0.09–0.61 µm) narrower retinal arterioles and 0.35 µm (−0.03 to 0.73 µm) wider venules. Children living 100 m closer to a major road had 0.30 µm (0.05–0.54 µm) narrower arterioles.Blood vessel diameters of the retinal microcirculation of healthy school-aged children respond to same-day PM2.5 exposure. Furthermore, children living closer to major roads had smaller arteriolar diameters. Our results suggest that the microcirculation, with retinal microvasculature as a proxy in this study, is a pathophysiological target for air pollution in children.

Electrocardiographic Imaging (ECGI) reconstructs heart surface potentials (HSPs) from body surface potentials (BSPs) using a patient-specific torso-heart geometry derived from CT or MRI. Potential inaccuracies in the estimate of the torso-heart geometry, and of the electrode positions on the body surface may limit the accuracy of the reconstructed HSPs. In this study, we aim at providing a proof-of-principle Deep Neural Network (DNN) which directly maps BSPs to HSPs, without the need to estimate a transfer matrix for the forward problem. In particular, we propose a torso normalization method that can turn the original torso geometry into a 2D image, and transform it into a normalized body surface potential map (nBSPM). In this paper, we did this for four dog geometries. A 4-layer back propagation (BP) neural network is built and trained with the nBSPMs as the input, and the pixel values of corresponding heart surface potential maps (HSPMs) as the output. The experiments show that the mean-squared error (MSE) on the validation data set decreases and converges to around 3.0. The value of the Pearson correlation coefficient between the original and reconstructed HSPMs on the test data set is larger than 0.92. This indicates that the proposed model is suitable for reconstructing HSPs from BSPs.

In contrast to electrograms, Body-Surface Potential Mapping (BSPM) records the global atrial activity, at the expenses of a lower spatial accuracy. The aim of this study is to investigate whether BSPM recordings can discriminate persistent patients undergoing electrical cardiover-sion, based on the body-surface normalized AF spatial frequency distribution. High-density BSPMs (120 anterior, 64 posterior electrodes) were recorded in 63 patients with persistent AF. For each patient and electrode recording, the frequency content of AF was analyzed on the raw signal, and also by means of the normalized correlation function, and by Singular Spectrum Analysis (SSA). In order to compare the body-surface spatial distributions of AF frequency in all patients, these distributions were first normalized, before performing statistical analysis. We found that the distribution of AF frequency on the body-surface, and its interpretation, are strongly dependent on the specific method employed. Moreover, the estimated body-surface AF frequency was greater over the central posterior and the right anterior BSPM locations. Finally, SSA-based decomposition followed by frequency analysis could discriminate AF patients recurring 4 to 6 weeks after electrical cardioversion from those who did not, based on the frequency content in the proximity of V1.

Atrial propagation patterns during atrial fibrillation (AF) can be characterized by a certain degree of recurrence (associated with different types of reentrant circuits that can drive the arrhythmia). In this study, we investigated this recurrent activity at the level of the body-surface, by measuring the level of similarity between pairs of consecutive atrial vectors. High-density body surface potential maps (120 anterior, 64 posterior electrodes) were recorded in 75 patients in persistent AF. For each patient, atrial vectors were created by taking the samples from all electrodes at each time instant. Similarity between consecutive vectors was measured in terms of the value of the cosine of the angle between two vectors. In all patients, the series of cosine values showed a quasi-periodic behavior, with atrial vectors alternating between phases of slow motion, and phases of fast motion. Moreover, the frequency of this behavior is about twice the AF dominant frequency, which suggests that within one AF cycle there are two phases of slow motion and two of fast motion, alternating. Finally, the amount of slow phases is positively correlated with a higher long-term recurrent behavior of the atrial propagation patterns. This seems to indicate that atrial vectors may provide a new way to noninvasively investigate atrial fibrillation dynamics.