A5075183301- Model Reduction and Neural Networks
- Cardiac electrophysiology and arrhythmias
- Probabilistic and Robust Engineering Design
- Cardiac Arrhythmias and Treatments
- Cardiovascular Function and Risk Factors
- Atrial Fibrillation Management and Outcomes
- Neural Networks and Applications
- Time Series Analysis and Forecasting
- ECG Monitoring and Analysis
- Elasticity and Material Modeling
- Control Systems and Identification
- Fault Detection and Control Systems
- Anomaly Detection Techniques and Applications
- Computer Graphics and Visualization Techniques
- Data Analysis with R
- COVID-19 epidemiological studies
- Advanced Control Systems Optimization
- Gaussian Processes and Bayesian Inference
- 3D Shape Modeling and Analysis
- EEG and Brain-Computer Interfaces
- Target Tracking and Data Fusion in Sensor Networks
- Neuroscience and Neural Engineering
- NMR spectroscopy and applications
- Force Microscopy Techniques and Applications
- Electron Spin Resonance Studies
Politecnico di Milano
2016-2024
École Polytechnique Fédérale de Lausanne
2020-2022
Abstract Predicting the evolution of systems with spatio-temporal dynamics in response to external stimuli is essential for scientific progress. Traditional equations-based approaches leverage first principles through numerical approximation differential equations, thus demanding extensive computational resources. In contrast, data-driven deep learning algorithms describe system low-dimensional spaces. We introduce an architecture, termed Latent Dynamics Network, capable uncovering intrinsic...
Mechano-electric feedbacks (MEFs), which model how mechanical stimuli are transduced into electrical signals, have received sparse investigation by considering electromechanical simulations in simplified scenarios. In this paper, we study the effects of different MEFs modeling choices for myocardial deformation and nonselective stretch-activated channels (SACs) monodomain equation. We perform numerical during ventricular tachycardia (VT) employing a biophysically detailed anatomically...
Computational inverse problems related to partial differential equations (PDEs) often contain nuisance parameters that cannot be effectively identified but still need considered as part of the problem. The objective this work is show how take advantage a reduced order framework speed up Bayesian inversion on identifiable system, while marginalizing away (potentially large number of) parameters. key ingredients are twofold. On one hand, we rely basis (RB) method, equipped with computable...
Simulating the cardiac function requires numerical solution of multi-physics and multi-scale mathematical models. This underscores need for streamlined, accurate, high-performance computational tools. Despite dedicated endeavors various research teams, comprehensive user-friendly software programs simulations, capable accurately replicating both normal pathological conditions, are still in process achieving full maturity within scientific community.This work introduces [Formula: see...
Abstract The development of biophysical models for clinical applications is rapidly advancing in the research community, thanks to their predictive nature and ability assist interpretation data. However, high-resolution accurate multi-physics computational are computationally expensive personalisation involves fine calibration a large number parameters, which may be space-dependent, challenging translation. In this work, we propose new approach, relies on combination physics-informed neural...
The ensemble Kalman filter is a computationally efficient technique for solving state and/or parameter estimation problems in the framework of statistical inversion when relying on Bayesian paradigm. Unfortunately, its cost may become moderately large systems described by nonlinear time-dependent PDEs, because entailed each PDE query. In this paper we propose reduced basis to address above problems. reduction stage yields intrinsic approximation errors, whose propagation through filtering...
Abstract We present a new, computationally efficient framework to perform forward uncertainty quantification (UQ) in cardiac electrophysiology. consider the monodomain model describe electrical activity tissue, coupled with Aliev‐Panfilov characterize ionic through cell membrane. address complete UQ pipeline, including both: (i) variance‐based global sensitivity analysis for selection of most relevant input parameters, and (ii) way propagation investigate impact intra‐subject variability on...
<abstract><p>The numerical simulation of several virtual scenarios arising in cardiac mechanics poses a computational challenge that can be alleviated if traditional full-order models (FOMs) are replaced by reduced order (ROMs). For example, the case problems involving vector input parameters related, e.g., to material coefficients, projection-based ROMs provide mathematically rigorous physics-driven surrogate ROMs. In this work we demonstrate how, once trained, yield extremely...
In the context of cardiac electrophysiology, we propose a novel computational approach to highlight and explain long-debated mechanisms behind atrial fibrillation (AF) reliably numerically predict its induction sustainment. A key role is played, in this respect, by new way setting parametrization electrophysiological mathematical models based on conduction velocities; these latter are estimated from high-density mapping data, which provide detailed characterization patients' substrate during...
In this work, we present a PDE-aware deep learning model for the numerical solution to inverse problem of electrocardiography. The both leverages data availability and exploits knowledge physically based mathematical model, expressed by means partial differential equations (PDEs), carry out task at hand. goal is estimate epicardial potential field from measurements electric discrete set points on body surface. employment techniques in context made difficult low amount clinical disposal, as...
In this paper, we propose a multi-fidelity approach for parameter estimation problems based on Physics-Informed Neural Networks (PINNs). The proposed methods apply to models expressed by linear or nonlinear differential equations, whose parameters need be estimated starting from (possibly partial and noisy) measurements of the model’s solution. To overcome limitations PINNs in case only few and/or significantly noisy data are available, train first Artificial Network (ANN), that provides...
Predicting the evolution of systems that exhibit spatio-temporal dynamics in response to external stimuli is a key enabling technology fostering scientific innovation. Traditional equations-based approaches leverage first principles yield predictions through numerical approximation high-dimensional differential equations, thus calling for large-scale parallel computing platforms and requiring large computational costs. Data-driven approaches, instead, enable description low-dimensional...
Electrogram (EGM) fractionation is often associated with diseased atrial tissue; however, mechanisms for occurring above an established threshold of 0.5 mV have never been characterized. We sought to investigate during sinus rhythm (SR) the underlying bipolar EGM high-density mapping in patients fibrillation (AF).Forty-five undergoing AF ablation (73% paroxysmal, 27% persistent) were mapped at high density (18562 ± 2551 points) SR (Rhythmia). Only EGMs voltages considered analysis. When (>...
We introduce universal solution manifold network (USM-Net), a novel surrogate model, based on artificial neural networks (ANNs), which applies to differential problems whose depends physical and geometrical parameters. employ mesh-less architecture, thus overcoming the limitations associated with image segmentation mesh generation required by traditional discretization methods. Our method encodes variability through scalar landmarks, such as coordinates of points interest. In biomedical...