Computer Methods and Programs in Biomedicine

Central venous pressure (CVP), a key component of hemodynamic monitoring, is widely used to guide fluid resuscitation in critically ill patients. It is typically measured using central venous line catheterization, which is the gold standard, but this method is invasive, time-consuming, and associated with complications. This study aims to investigate whether machine learning (ML)-based analysis of features extracted from a non-invasive, standard-of-care waveform--the photoplethysmography (PPG) signal--can identify patients with elevated CVP. We trained Light Gradient-Boosting Machine (LightGBM) model using a large perioperative dataset (MLORD), containing 17,327 surgical patients from 2019 to 2022 at UCLA. For this study, we selected 1665 patients with both PPG and CVP waveforms available. A total of 843 PPG features per cardiac cycle (CC) were extracted from the PPG waveforms using a signal processing-based feature extraction tool, along with the simultaneous maximum value calculated from the corresponding CCs in the CVP waveform. Additionally, for each patient, the average and standard deviation of each PPG feature, as well as the mean of the maximum CVP values, were calculated across all cardiac cycles, resulting in 843 averaged PPG features, 843 PPG feature standard deviations, and one averaged maximum CVP value per patient. The average maximum CVP value was used as the ground truth to classify patients as either normal (5 [≤] CVP [≤] 15 mmHg) or elevated (CVP > 15 mmHg). Of the 1,665 patients, 1,182 were normal and 483 were elevated. The dataset was split into 90% for training (1,063 normal and 435 elevated) and 10% for testing (119 normal and 48 elevated). From the 1686 PPG features (843 averaged and 843 standard deviation), 246 were selected for model development using the Recursive Feature Elimination with Cross-Validation (RFECV) approach. To further enhance performance, hyperparameters were tuned through 5-fold cross-validation on the training set. Finally, the best-performing configuration was retrained on the full training data, and its performance was evaluated on the held-out test set. To provide a robust estimate and confidence interval, a bootstrapping procedure with 100 iterations was performed on the test set. The LightGBM classifier achieved a mean area under the receiver operating characteristic curve (AUC) of 0.79 (95% CI: 0.71-0.84) and mean accuracy of 0.71 (95% CI: 0.65-0.77), demonstrating good discriminatory power in distinguishing between patients with normal and elevated CVP. This study highlights the ability of PPG-derived features to discriminate between patients with normal and elevated CVP using ML. These early findings lay the groundwork for future research aimed at developing non-invasive approaches to CVP assessment.

Arterial blood pressure (ABP) and photoplethysmography (PPG) waveforms both contain vital physiological information for the prevention and treatment of cardiovascular diseases. Extracted features from these waveforms have diverse clinical applications, including predicting hyper- and hypo-tension, estimating cardiac output from ABP, and monitoring blood pressure and nociception from PPG. However, the lack of standardized tools for feature extraction limits their exploration and clinical utilization. In this study, we propose an automatic feature extraction tool that first detects temporal location of landmarks within each cardiac cycle of ABP and PPG waveforms, including the systolic phase onset, systolic phase peak, dicrotic notch, and diastolic phase peak using the iterative envelope mean method. Then, based on these landmarks, extracts 852 features per cardiac cycle, encompassing time-, statistical-, and frequency-domains. The tools ability to detect landmarks was evaluated using ABP and PPG waveforms from a large perioperative dataset (MLORD dataset) comprising 17,327 patients. We analyzed 34,267 cardiac cycles of ABP waveforms and 33,792 cardiac cycles of PPG waveforms. Additionally, to assess the tools real-time landmark detection capability, we retrospectively analyzed 3,000 cardiac cycles of both ABP and PPG waveforms, collected from a Philips IntelliVue MX800 patient monitor. The tools detection performance was assessed against markings by an experienced researcher, achieving average F1-scores and error rates for ABP and PPG as follows: (1) On MLORD dataset: systolic phase onset (99.77 %, 0.35 % and 99.52 %, 0.75 %), systolic phase peak (99.80 %, 0.30 % and 99.56 %, 0.70 %), dicrotic notch (98.24 %, 2.63 % and 98.72 %, 1.96 %), and diastolic phase peak (98.59 %, 2.11 % and 98.88 %, 1.73 %); (2) On real time data: systolic phase onset (98.18 %, 3.03 % and 97.94 %, 3.43 %), systolic phase peak (98.22 %, 2.97 % and 97.74 %, 3.77 %), dicrotic notch (97.72 %, 3.80 % and 98.16 %, 3.07 %), and diastolic phase peak (98.04 %, 3.27 % and 98.08 %, 3.20 %). This tool has significant potential for supporting clinical utilization of ABP and PPG waveform features and for facilitating feature-based machine learning models for various clinical applications where features derived from these waveforms play a critical role.

Arterial blood pressure (ABP) and photoplethysmography (PPG) waveforms contain valuable clinical information and play a crucial role in cardiovascular health monitoring, medical research, and managing medical conditions. The features extracted from PPG waveforms have various clinical applications ranging from blood pressure monitoring to nociception monitoring, while features from ABP waveforms can be used to calculate cardiac output and predict hypertension or hypotension. In recent years, many machine learning models have been proposed to utilize both PPG and ABP waveform features for these healthcare applications. However, the lack of standardized tools for extracting features from these waveforms could potentially affect their clinical effectiveness. In this paper, we propose an automatic signal processing tool for extracting features from ABP and PPG waveforms. Additionally, we generated a PPG feature library from a large perioperative dataset comprising 17,327 patients using the proposed tool. This PPG feature library can be used to explore the potential of these extracted features to develop machine learning models for non-invasive blood pressure estimation.

Reproducibility SummaryO_ST_ABSScope of ReproducibilityC_ST_ABSThe authors of the original paper present six benchmark tasks on the previously published PTB-XL dataset, containing, 21837 12-lead ECGs from 18885 patients. They evaluate seven different neural network architectures on the six bench-mark tasks. The authors have published all code and claim full reproducibility. In addition, they published code for easy implementation of new models. To validate the claim of reproducibility we implemented a new model and tested it, and the seven models presented by the authors of the original paper, on the six benchmark tasks. MethodologyWe used the publicly available code, published by the authors of the original paper, as a starting point for our experiment. Furthermore, we modified the code slightly in order to make it compatible with a cloud-hosted Jupyter Notebook, Google Colab. We ran the experiments using Google Colab Pro, using 32 GB RAM and either 1 x NVIDIA P100 or 1 x NVIDIA T4 GPU. ResultsWe successfully managed to reproduce the original work and also verified the validity of the main claims of the original paper. In addition, we showed how robust the models were to noise and finally implemented a new model that showed comparable performance with the models proposed in the original paper. What was easyThe publicly available code published by the authors made it easy to reproduce and obtain the same results as reported in their paper. What was difficultWe faced two main issues in this work. (1) running the code in a cloudhosted jupyter notebook. This was done in order to get access to free or cheap GPUs. (2) Implement own models using the provided template. The description on how to use the base class and the configuration file could have been more detailed. Communication with original authorsCommunication with the authors of the original paper was established early in the project and helped us by clarifying some aspects of the work. In the final stage of this project the authors of the original paper were given this manuscript in order to read it and provide feedback.

One goal of Scientific Machine Learning (SciML) is to advance traditional scientific computing frameworks with modern machine learning tools. This includes extending established methods, such as the finite element method, with cardiac function applications due to their complexity and need for very rapid execution times for real time clinical use. In this work, we present an advanced form of the Neural Network Finite Element (NNFE) method specialized for cardiac simulations, termed CARDIAX-NNFE. The NNFE method learns the parameter-to-displacement field map by training over the residual of the hyperelastic material PDE, using the domain represented by finite elements. The implementation is developed in Python using JAX to leverage its automatic differentiation, highly parallel GPU, and JIT-compilation capabilities. To demonstrate CARDIAX-NNFE effectiveness, we trained full cardiac pressure-volume responses using a simplified heart model, spanning the entire cardiac physiological functional range. Results indicated the ability to simulate a family of pressure-volume solutions with average nodal positional error of 0.023 mm and maximal error of 0.054 mm, with a single complete PV loop evaluated in 0.002 seconds. The CARDIAX-NNFE software platform thus provides for a robust platform for cardiac functional simulations. Moreover, it provides the structure for residual-based SciML methods, which can apply to a variety of physics-based biomedical problems that require high execution speed for clinical applications.

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSAimsC_ST_ABSWe aim to develop a patient-specific computational model to predict the risk of Ventricular Tachycardia (VT) in patients with Biventricular Cardiac Resynchronization Therapy (BiV-CRT) device. Patients are indeed at risk of developing arrhythmias due to BiV-CRT pacing, a known potential complication that puts the cardiologist on guard against its prevention. Materials and MethodsWe consider three non-ischemic fibrotic patients. Patient-specific left ventricle geometries and fibrosis regions are extracted from Cine-MRI and LGE-MRI. The electrophysiology model, based on the monodomain equation and on the Ten Tusscher-Panfilov (TTP06) ionic current model, is personalized using pre-operative Electro-Anatomical Mappings System data. The TTP06 parameters are adapted to reflect the altered electrical properties of the fibrotic tissue. To test inducibility, we use an [S]1-[S]2 stimulation protocol: [S]1 simulates the clinical BiV-CRT pacing with patient-specific VV-delay, followed by a [S]2 ectopic impulse. This procedure is repeated for ten ectopic sites. The arrhytmogenic risk is quantified by the number of ectopic sites that successfully generates a reentry loop. ConclusionThe models prediction of VT risk is consistent with the long-term clinical follow-up for all the patients. Arrhythmic patients show a higher number of ectopic sites from which a reentry loop is generated compared to the non-arrhythmic patient. This study provides a first, preliminary attempt towards the use of computational tools in assessing the vulnerability of the arrhythmic substrate during BiV-CRT pacing in non-ischemic patients. In future, such tools could serve as a powerful non-invasive diagnostic metric to inform clinicians about possible therapies to associate to BiV-CRT.

The analysis of electrophysiological signals from the human body has become increasingly crucial, especially given the widespread adoption of wearable technologies and the growing trend of remote and online monitoring. In situations where demographic patient data is unavailable, the evaluation of such information from electrophysiological signals becomes imperative for making well-informed diagnostic and therapeutic decisions, particularly in ambulatory and urgent cases. This study underscores the significance of this necessity by utilizing intracardiac electrograms to predict patient weight. Intracardiac electrograms were recorded from 44 patients (14 female, with an average age of 59.2{+/-}11.5 years) using a 64-pole basket catheter over a duration of 60 seconds. A dataset comprising 2,816 unipolar electrogram signal segments, each lasting 4 seconds, was utilized. Weight, considered as a continuous variable, underwent discretization into k bins with uniformly distributed widths, where various values of k were experimented with. As the value of k increases, class imbalance also increases. The state-of-the-art time series classification algorithm, Minirocket, was employed alongside the popular machine learning algorithm eXtreme Gradient Boosting (XGBoost). Minirocket consistently demonstrates superior performance compared to XGBoost across all class number scenarios and across all evaluation metrics, such as accuracy, F1 score, and Area Under the Curve (AUC) values, achieving scores of approximately 0.96. Conversely, XGBoost shows signs of overfitting, particularly noticeable in scenarios with higher class imbalance. Tuning probability thresholds for classes could potentially mitigate this issue. Additionally, XGBoosts performance improves with reduced bin numbers, emphasizing the importance of balanced classes. This study provides novel insights into the predictive capabilities of these algorithms and their implications for personalized medicine and remote health monitoring.

Accurate assessment of consciousness during general anesthesia is crucial for optimizing anesthetic dosage and patient safety. Current electroencephalogram-based monitoring devices can be inaccurate or unreliable in specific surgical contexts (e.g. cephalic procedures). This study investigated the feasibility of using electrocardiogram (ECG) features and machine learning to differentiate between awake and anesthetized states. A cohort of 48 patients undergoing surgery under general anesthesia at the Tours hospital was recruited. ECG-derived features were extracted, including spectral power, heart rate variability and complexity metrics, as well as heart rate fragmentation indices (HRF). These features were augmented by a range of physiological variables. The aim was to evaluate a number of machine learning algorithms in order to identify the most appropriate method for classifying the awake and anesthetized states. The gradient boosting algorithm achieved the highest accuracy (0.84). Notably, HRF metrics exhibited the strongest predictive power across all models tested. The generalizability of this ECG-based approach was further assessed using public datasets (VitalDB, Fantasia, and MIT-BIH Polysomnographic), achieving accuracies above 0.80. This study provides evidence that ECG-based classification methods can effectively distinguish awake from anesthetized states, with HRF indices playing a pivotal role in this classification. Author summaryGeneral anesthesia monitoring is critical for optimizing patient safety and outcomes. While electroencephalogram (EEG)-based systems are commonly used, they have limitations in accuracy and applicability, particularly in cases where EEG electrodes placement is challenging or impossible, such as during cephalic surgeries or when patients have forehead skin lesions. Here, a novel approach using electrocardiogram (ECG) signals and machine learning techniques was used to differentiate between awake and anesthetized states during surgery. A total of 48 patients undergoing surgical procedures under general anaesthesia at the Tours hospital were selected for inclusion in the study. This investigation focused on heart rate fragmentation indices, metrics designed for assessing biological versus chronological age, derived from ECG recordings. The gradient boosting algorithm demonstrates performance comparable to leading methods reported in the literature for this classification task. Importantly, model generalizability was confirm through successful application to publicly available datasets. This article highlights the potential of ECG signals as an alternative source for deriving depth of anesthesia indices, offering increased versatility in clinical settings where EEG monitoring is challenging or contraindicated.

BackgroundTransesophageal echocardiography (TEE) is a critical tool in diagnosing and managing infectious endocarditis, providing detailed images of cardiac structures. However, identifying vegetations on valves and their dynamic behavior in ultrasound videos can be challenging. TEEs metadata often does not include scale enabling computation of speed. ObjectivesTo address this, we developed a simple Python-based tool that enhances the visualization of these dynamic characteristics. This tool reconstructs an optical flow from TEE images, capturing the motion of cardiac structures and offering deeper insights into their behavior. The tool also recovers scale from visual information on the TEES. MethodsBy leveraging the Marching Cubes algorithm and 2D Fast Fourier Transform (FFT) to recover scale from images, the tool efficiently processes video frames to create a 3D representation where time is the third dimension. Wit his mouse the user can select temporal slices and a view of the dynamic evolution in that slice is created together with the speeds. ResultsThis approach allows for measurement of thicknesses and speeds, aiding in the evaluation of valvular and vegetation dynamics. ConclusionsThe tools user-friendly interface, built with Dash and Plotly, enables interactive analysis and visualization, making it a valuable asset for cardiologists in clinical settings to further analyze valvular behavior.

Electrocardiographic imaging (ECGI) presents a clinical opportunity to noninvasively understand the sources of arrhythmias for individual patients. To help increase the effectiveness of ECGI, we provide new ways to visualize associated measurement and modeling errors. In this paper, we study source localization uncertainty in two steps: First, we perform Monte Carlo simulations of a simple inverse ECGI source localization model with error sampling to understand the variations in ECGI solutions. Second, we present multiple visualization techniques, including confidence maps, level-sets, and topology-based visualizations, to better understand uncertainty in source localization. Our approach offers a new way to study uncertainty in the ECGI pipeline.

This study presents a novel two-stage framework to enhance the reliability of resting electrocardiogram (ECG) signals by addressing motion artifacts that often compromise diagnostic accuracy. In the first stage, motion artifacts are mitigated using stationary wavelet transform coupled with Savitzky-Golay filtering, effectively preserving critical ECG morphological features such as the QRS complex. The second stage employs a deep convolutional neural network to classify ECG signals as either usable or artifact-corrupted, achieving a classification accuracy of 98.76%. Utilizing a 12-lead ECG dataset from PhysioNet, the proposed unified CNN model outperforms individual lead-specific models, offering superior computational efficiency (1.6 seconds vs. 21.7 seconds for predictions) and reduced storage requirements (1 GB vs. 15 GB). The approach demonstrates high sensitivity (98.74%) and specificity (98.77%), ensuring robust detection of noisy signals. By integrating advanced preprocessing with deep learning, this framework enhances ECG signal clarity, reducing the risk of misdiagnosis in clinical settings.

Transcatheter aortic valve replacement (TAVR) has rapidly displaced surgical aortic valve replacement (SAVR). However, certain post-TAVR complications persist, with cardiac conduction abnormalities (CCA) being one of the major ones. The elevated pressure exerted by the TAVR stent onto the conduction fibers situated between the aortic annulus and the His bundle, in proximity to the atrioventricular (AV) node, may disrupt the cardiac conduction leading to the emergence of CCA. In his study, an in-silico framework was developed to assess the CCA risk, incorporating the effect of a dynamic beating heart and pre-procedural parameters such as implantation depth and preexisting cardiac asynchrony in the new onset of post-TAVR CCA. A self-expandable TAVR device deployment was simulated inside an electro-mechanically coupled beating heart model in five patient scenarios, including three implantation depths, and two preexisting cardiac asynchronies: (i) a right bundle branch block (RBBB) and (ii) a left bundle branch block (LBBB). Subsequently, several biomechanical parameters were analyzed to assess the post-TAVR CCA risk. The results manifested a lower cumulative contact pressure on the conduction fibers following TAVR for aortic deployment (0.018 MPa) compared to baseline (0.29 MPa) and ventricular deployment (0.52 MPa). Notably, the preexisting RBBB demonstrated a higher cumulative contact pressure (0.34 MPa) compared to the baseline and preexisting LBBB (0.25 MPa). Deeper implantation and preexisting RBBB cause higher stresses and contact pressure on the conduction fibers leading to an increased risk of post-TAVR CCA. Conversely, implantation above the MS landmark and preexisting LBBB reduces the risk.

Here we use a statistical atlas of end-diastolic (ED) and end-systolic (ES) biventricular shapes - previously derived from the UK Biobank imaging substudy - to generate meshes for finite element (FE) simulations of ventricular wall mechanics. The models used the Holzapfel-Ogden constitutive law for passive material properties and a time-varying elastance model of systolic tension development. Simulated ED and ES deformations were projected onto the shape atlas and the principal components were used to train a multi-layer perceptron as a surrogate model. The input layer included shape modes of the unloaded ventricular geometry, and material parameters and ventricular pressures at ED and ES. After training with 444 simulations, the surrogate model achieved a mean square error in predicted displacements of < 2 mm and volumetric overlaps with FE-predicted deformed shapes > 97%, demonstrating good fidelity to the simulated ground truth. This approach may enable accurate prediction of ventricular wall mechanics without computationally expensive finite element analysis, offering a more feasible method for rapid, subject-specific cardiac modeling.

We present a novel methodology that generates a three-dimensional (3D) electrocardiogram (ECG) directly derived from standard two-dimensional (2D) recordings using conventional cardiac electrode configurations. Through spherical-to-Cartesian coordinate transformations, we generate 3D representations that form loops over the standard deflections used in routine medical practice, while geometrically integrating time with voltage within the coordinates themselves. To validate our approach from a clinical perspective, we analyzed datasets focused on acute myocardial ischemia. We evaluated the diagnostic value by comparing 2D and 3D ECG metrics, specifically perimeter and curvature, across different ischemic states. Additionally, we propose a novel mathematical formulation, analogous to curvature, designed to more accurately detect variations in the progression of ischemia. This formulation, referred to as "almost-curvature", achieves maximum efficiency when integrated with our 3D representation method. Our results highlighted significant geometric differences in the 3D metrics, demonstrating their potential to detect ischemic alterations with greater effectiveness than traditional methods. These findings support the potential of 3D ECG as a transformative tool in cardiac diagnostics and underscore the need for continued research to expand its clinical applicability.

This study investigates Kolmogorov-Arnold Networks (KANs) for biosignal analysis, using electrocardiogram signals as a case study. KANs provide flexibility and require few parameters, making them suitable for wearable and edge devices. A simple KAN model with one hidden layer of 64 neurons was trained on the TNMG dataset and tested on the CPSC 2018 dataset, achieving an F1-score of 0.75 and AUROC of 0.95 on TNMG, and an F1-score of 0.62 and AUROC of 0.84 on CPSC. The model also showed robustness to missing channels, maintaining reasonable performance with only a single ECG lead. Compared with traditional Multi-Layer Perceptrons (MLPs) and Neural Circuit Policies (NCPs), KANs demonstrated greater flexibility, adaptability, interpretability, and efficiency. Additionally, a shallow network (CKAN) that integrates a single Conv2dLSTM layer with a small set of KAN neurons, mirroring two architectures built with different NCP neurons for TinyML, achieved an F1-score of 0.84 and an AUROC of 0.97 on TNMG, and an F1-score of 0.72 and an AUROC of 0.92 on CPSC. Incorporating learnable sparsity, a key feature of NCP neurons, into KAN neurons surprisingly enhanced both performance and generalization. Even after pruning sparse weights, the model maintained strong performance, surpassing the counterpart without sparsity.

BackgroundThree-dimensional (3D) electrocardiography (ECG) is a recent methodological advance that extends the dimensionality of the standard ECG, enabling geometric descriptors that capture acute ischemia. Integrating these descriptors with deep learning (DL) may improve the discrimination between ischemic and non-ischemic states and promote the clinical translation of 3D ECG analysis. MethodsECGs from seventeen patients with acute left anterior descending (LAD) artery stenosis (>50 %) were obtained from the PTB Diagnostic ECG Database (PhysioNet). Pre- and post-catheterization recordings were analyzed in 2D and 3D (V3, V6, time) over the QRS end-T onset interval. Geometric descriptors included perimeter, curvature, three almost-curvature variants, and a newly defined torsion metric. Statistical analyses comprised univariate, bivariate, and multivariate tests (PERMANOVA), complemented by DL classification using a residual multilayer perceptron with patient-wise cross-validation, isotonic calibration, and logistic meta-blending, adopting a significance level of = 0.01 (99 % confidence) to ensure inference stability given the limited sample size. ResultsFour descriptors changed significantly after revascularization (P2D,V 6t,{kappa} 2D,V 6t, 3D,2, and{tau} ). Correlation analyses indicated redundancy among curvature-related metrics, whereas torsion provided independent information. PERMANOVA confirmed that torsion alone, and only metric sets including torsion, achieved significance (p < 0.05). The torsion-based DL model provided the best discrimination, with an area under the ROC curve of 0.76 (99 % CI, 0.57-0.94; p < 0.001), specificity 0.82, and a Brier score of 0.18. ConclusionsThe integration of torsion into a DL-based 3D ECG framework enhanced the detection of acute ischemia, increasing diagnostic specificity and improving early triage and clinical decision-making in acute cardiac care.

The application of artificial intelligence (AI) in the medical field has seen a significant increase in popularity, particularly for its ability to accurately detect abnormalities across a range of diagnostic tests. The effectiveness and precision of AI models are highly contingent on the quality and diversity of the training data used in their development. In the present work, we have developed an open-source AI model designed to anonymize electrocardiogram (ECG) recordings. This model achieves anonymization by automatically detecting and extracting the waveform data. This tool can be used to prepare input data that in turn serve as input variables for training AI models specifically for cardiology applications. By ensuring that patient-identifying information is removed while retaining the essential waveform data. The present model facilitates the creation of robust, privacy-preserving datasets that can enhance the training and performance of AI in cardiology.

BackgroundThe MIT-BIH Noise Stress Test Database (NSTDB) is a publicly available resource for testing QRS detection algorithms. Serial QRS detection algorithms applied to the NSTDB have apparently failed to detect the presence of a possible heartbeat like rhythm associated with peaks that the NSTDB classifies as noise. The failure to detect this rhythm may arise from the difficulty associated with interpreting noisy RR interval time series produced by serial QRS detection schemes. Algorithm Summary and ExperimentTo extract rhythm information from noisy peak time/RR interval time series, a peak space signal is created with triangular pulses centered on peaks located by a serial QRS detection algorithm such as Pan-Tompkins. The peak space signal is autocorrelated over 20 s segments and the primary (non-origin) peak in the autocorrelation signal is located. In the presence of reasonably regular sinus rhythm, this peak corresponds to a fundamental RR interval present throughout the 20 s segment. This peak time processing method was applied to the Pan-Tompkins QRS detections in the motion artifact record of the NSTDB. To compare the results to a different algorithm capable of detecting patterns at the segment level, a previously described pattern-based heartbeat detection scheme (Temporal Pattern Search, or "TEPS") was applied in both single and multiple channel modes to the NSTDB motion artifact record. ResultsBoth the Pan-Tompkins/autocorrelation method and TEPS detected a persistent rhythm around 1000-1050 ms in both channels throughout the entirety of the motion artifact record. The RR interval correlation between Pan-Tompkins/autocorrelation and single channel TEPS was 0.8 and 0.7 in channels 1 and 2 respectively with p values of 0.

AO_SCPLOWBSTRACTC_SCPLOWThe early phase of acute myocardial ischemia is associated with an elevated risk of ventricular reentrant arrhythmias. Indeed, after partial or total occlusion of a coronary artery, some regions of the heart experience a reduction in myocardial blood flow. This causes metabolic and cellular processes, such as hypoxia, hyperkalemia and acidosis, which lead to changes in the transmembrane ionic dynamics. The effect of such alterations may result in the formation of electrical loops and reentries. Computational models could simulate the generation of arrhythmias, possibly persistent, in condition of ectopic beats and in presence of acute myocardial regions. Since quantitative information (extent, localization, ...) about acute ischemic regions are hardly available from clinics, to date, computational models only integrate imaging data from chronic infarcted ventricles. This may not accurately reflect the acute condition. This work presents a novel patient-specific electrophysiological model, based on images of myocardial blood flow maps acquired during a pharmacologically induced acute ischemic event. The model personalization is obtained with the partitioning of the left ventricle geometries on the basis of the myocardial blood flow maps. First, we aim to numerically investigate the induction and sustainment of reentrant drivers in patient-specific scenarios, in order to assess their arrhythmic propensity. Secondly, we perform an intra-patient sensitivity analysis, where different levels of acute ischemia are virtually depicted for the most arrhythmogenic patient. Our results suggest that the amount of ischemic regions seems to have less influence on arrhythmogenesis than their pattern.

In this study, the performances of 33 supervised and unsupervised machine learning methods for the automatic classification of cardiac arrhythmias as normal or abnormal using the MIT BIH Arrhythmia Database are evaluated. Electrocardiogram signals from the MLII and V1 leads are segmented into fixed-length windows aligned to the R peak, with raw amplitude values used as model inputs without feature extraction. In the supervised analysis, various statistical and ensemble classifiers are evaluated, while the unsupervised analysis assesses Isolation Forest, One Class support vector machines (SVM), Local Outlier Factor, Elliptic Envelope, and an autoencoder model. The results demonstrate that, when labeled data are available, supervised methods, particularly K nearest neighbors (KNN) and Random Forest, provide higher accuracy and more balanced classification compared with unsupervised models. Unsupervised approaches, on the other hand, are positioned as complementary tools for arrhythmia screening and early warning when labeled data are limited.