arXiv subjects

Natalia A. Trayanova

Publications and source records attributed to Natalia A. Trayanova.

4 recordsLinked to original sources

arXiv2025

A computational pipeline for clustering left atrial appendage morphology via elastic shape analysis

Morphological variations in the left atrial appendage (LAA) are associated with different levels of ischemic stroke risk for patients with atrial fibrillation (AF). Studying LAA morphology can elucidate mechanisms behind this association and lead to the development of advanced stroke risk stratification tools. However, current categorical descriptions of LAA morphologies are qualitative and inconsistent across studies, which impedes advancements in our understanding of stroke pathogenesis in AF. To mitigate these issues, we introduce a quantitative pipeline that combines elastic shape analysis with unsupervised learning for the categorization of LAA morphology in AF patients. As part of our pipeline, we compute pairwise elastic distances between LAA meshes from a cohort of 20 AF patients, and leverage these distances to cluster our shape data. We demonstrate that our method clusters LAA morphologies based on distinctive shape features, overcoming the innate inconsistencies of current LAA categorization systems, and paving the way for improved stroke risk metrics using objective LAA shape groups.

q-bio.QM↗

arXiv2024

Defining myocardial fiber bundle architecture in atrial digital twins

A key component in developing atrial digital twins (ADT) - virtual representations of patients' atria - is the accurate prescription of myocardial fibers which are essential for the tissue characterization. Due to the difficulty of reconstructing atrial fibers from medical imaging, a widely used strategy for fiber generation in ADT relies on mathematical models. Existing methodologies utilze semi-automatic approaches, are tailored to specific morphologies, and lack rigorous validation against imaging fiber data. In this study, we introduce a novel atrial Laplace-Dirichlet-Rule-Based Method (LDRBM) for prescribing highly detailed myofiber orientations and providing robust regional annotation in bi-atrial morphologies of any complexity. The robustness of our approach is verified in eight extremely detailed bi-atrial geometries, derived from a sub-millimiter Diffusion-Tensor-Magnetic-Resonance Imaging (DTMRI) human atrial fiber dataset. We validate the LDRBM by quantitatively recreating each of the DTMRI fiber architectures: a comprehensive comparison with DTMRI ground truth data is conducted, investigating differences between electrophysiology (EP) simulations provided by either LDRBM and DTMRI fibers. Finally, we demonstrate that the novel LDRBM outperforms current state-of-the-art fiber models, confirming the exceptional accuracy of our methodology and the critical importance of incorporating detailed fiber orientations in EP simulations. Ultimately, this work represents a fundamental step toward the development of physics-based digital twins of the human atria, establishing a new standard for prescribing fibers in ADT.

physics.med-ph↗

arXiv2021

Fast Posterior Estimation of Cardiac Electrophysiological Model Parameters via Bayesian Active Learning

Probabilistic estimation of cardiac electrophysiological model parameters serves an important step towards model personalization and uncertain quantification. The expensive computation associated with these model simulations, however, makes direct Markov Chain Monte Carlo (MCMC) sampling of the posterior probability density function (pdf) of model parameters computationally intensive. Approximated posterior pdfs resulting from replacing the simulation model with a computationally efficient surrogate, on the other hand, have seen limited accuracy. In this paper, we present a Bayesian active learning method to directly approximate the posterior pdf function of cardiac model parameters, in which we intelligently select training points to query the simulation model in order to learn the posterior pdf using a small number of samples. We integrate a generative model into Bayesian active learning to allow approximating posterior pdf of high-dimensional model parameters at the resolution of the cardiac mesh. We further introduce new acquisition functions to focus the selection of training points on better approximating the shape rather than the modes of the posterior pdf of interest. We evaluated the presented method in estimating tissue excitability in a 3D cardiac electrophysiological model in a range of synthetic and real-data experiments. We demonstrated its improved accuracy in approximating the posterior pdf compared to Bayesian active learning using regular acquisition functions, and substantially reduced computational cost in comparison to existing standard or accelerated MCMC sampling.

cs.LG↗

arXiv2021

Anatomically-Informed Deep Learning on Contrast-Enhanced Cardiac MRI for Scar Segmentation and Clinical Feature Extraction

Visualizing disease-induced scarring and fibrosis in the heart on cardiac magnetic resonance (CMR) imaging with contrast enhancement (LGE) is paramount in characterizing disease progression and quantifying pathophysiological substrates of arrhythmias. However, segmentation and scar/fibrosis identification from LGE-CMR is an intensive manual process prone to large inter-observer variability. Here, we present a novel fully-automated anatomically-informed deep learning solution for left ventricle (LV) and scar/fibrosis segmentation and clinical feature extraction from LGE-CMR. The technology involves three cascading convolutional neural networks that segment myocardium and scar/fibrosis from raw LGE-CMR images and constrain these segmentations within anatomical guidelines, thus facilitating seamless derivation of clinically-significant parameters. In addition to available LGE-CMR images, training used "LGE-like" synthetically enhanced cine scans. Results show excellent agreement with those of trained experts in terms of segmentation (balanced accuracy of $96\%$ and $75\%$ for LV and scar segmentation), clinical features ($2\%$ difference in mean scar-to-LV wall volume fraction), and anatomical fidelity. Our segmentation technology is extendable to other computer vision medical applications and to problems requiring guidelines adherence of predicted outputs.

eess.IV↗