What's NewOur PeopleGallery/ProjectsNews ArchivePublications

AVERT-VT: Ablation at Virtual-hEart pRedicted Targets for VT

We demonstrated the utility of virtual-heart simulations in determining noninvasively the optimal ventricular tachycardia (VT) ablation targets and guiding the clinical procedure of VT ablation (https://doi.org/10.1038/s41551-018-0282-2). The non-invasive approach was termed VAAT (virtual-heart arrhythmia ablation targeting) and was assessed in retrospective studies, and in a proof-of-concept prospective study.

We now proceed with AVERT-VT; FDA-approved clinical trial for 10 prospective patients to demonstrate decreased procedural times, decreased patient risk, and improved efficacy for the catheter ablation management of infarct-related VT. Personalized 3D ventricular models are reconstructed from patient’s cardiac LGE-MRI imaging data. VAAT protocol is repeated until all in silico VTs are terminated by the virtual lesions and complete VT non‐inducibilityis achieved. The set of ablation targets that achieve this are the final (optimal) set of VAAT ablation targets; representing the targets that would be directly approached during the clinical procedure, without any electrical mapping (https://doi.org/10.1002/wsbm.1477).

‍

Characterization of the Electrophysiologic Remodeling of Patients with Ischemic Cardiomyopathy

Patients with ischemic cardiomyopathy (ICMP) are at high risk for malignant arrhythmias, largely due to electrophysiological remodeling of the non-infarcted myocardium; however, little is known about the electrophysiological properties of the non-infarcted myocardium. To assess the pro-arrhythmic behavior of non-infarcted myocardium in ICMP patients, we coupled computational simulations with a genetic algorithm to establish a methodology for the development of disease-specific action potential models based on clinically-measured action potential duration restitution data. Patients with ICMP in addition to a steeper action potential restitution curve, have a greater spatial heterogeneity in the action potential restitution slope.

Personalized Hemodynamic Simulations to Understand Increased Stroke Risk in AF Patients with High Left Atrial Fibrotic Burden

Presence of left atrial (LA) fibrosis burden increases the stroke risk in atrial fibrillation (AF) patients. However, exact reasons as of how the increased LA fibrotic burden increases stroke risk remains unknown. We hypothesize that presence of fibrosis could lead to aberrant hemodynamics in the LA of AF patients, which could explain their increased stroke risk. We used personalized LA hemodynamic simulations on LGE-MRI images of AF patients to explore flow-dynamics in the fibrotic and compared to the non-fibrotic region at the LA wall. Preliminary results suggest that flow in the LA near the fibrotic region is aberrant, potentially leading to thrombogenesis and increasing the risk of stroke

Real-Time Prediction of Cardiovascular Complications in Hospitalized Patients with COVID-19

In a study funded by the NSF RAPID COVID program, we developed a machine learning model to provide real-time, continuously updating risk predictions for multiple adverse cardiovascular outcomes. The results of this study are available as a pre-print here: https://www.medrxiv.org/content/10.1101/2021.01.03.21249182v2‍

The COVID-HEART predictor was developed with millions of data points collected from 2,178 patients treated at the five hospitals in the Johns Hopkins Health System between March 1 and Sept. 27. It can predict cardiac arrest in COVID patients with a median early warning time of 18 hours, and can predict deadly blood clots that can travel to the lungs or brain with a median early warning time of 72 hours.

New Intraprocedural Automated System for Localizing Idiopathic Ventricular Arrhythmia Origins

Prior site of origin systems to identify idiopathic ventricular arrhythmias (IVA) are limited by the need to create complete electroanatomical maps (EAM), inability to localize intracardiac structures/vessels, and require pre-procedural cardiac imaging. Our Automatic Arrhythmia Origin Localization (AAOL) system addresses these issues. The AAOL system combines 3-lead, 120-ms QRS integrals with pace mapping to predict the site of earliest ventricular activation and project that site onto patient-specific EAM geometry. In a prospective, multicenter study of patients undergoing IVA catheter ablation, twenty-three IVA origin sites were localized by the AAOL system with a mean localization accuracy of 3.6 mm, better than any prior published system.

Feasibility Study Shows Concordance Between Image-based Virtual-Heart Ablation Targets and Predicted ECG-based Arrhythmia Exit-Sites

We recently developed two non-invasive methodologies to help guide VT ablation: population-derived automated VT exit localization (PAVEL) and virtual-heart arrhythmia ablation targeting (VAAT). We hypothesized that while very different in their nature, limitations, and type of ablation targets (substrate-based vs. clinical VT), the image-based VAAT and the ECG-based PAVEL technologies would be spatially concordant in their predictions. This project tests this hypothesis in ischemic cardiomyopathy patients in a retrospective feasibility study.

‍

Decreasing Overfitting and Increasing Generalizability of Deep Learning on Heart Rhythm Classification

Deep learning (DL) has achieved promising performance in detecting common heart rhythms from the 12-lead electrocardiogram (ECG). However, diagnostic redundancy exists in the 12-lead ECG, which could impose a systematic overfitting on DL, causing poor generalization. In this study, we developed and evaluated a novel multi-stage DL-based model that incorporated an ECG-lead subset selection module for automatic heart-rhythm classification. The ECG-lead subset selection module was introduced within the proposed model to efficiently constrain model complexity. Results demonstrated the efficacy of the subset selection approach and the feasibility of representing a complete 12-lead ECG by an optimal 4-lead subset (leads II, aVR, V1, V4) to improve DL models’ generalizability in the heart-rhythm classification.

Anatomically-Informed Deep Learning to Segment Contrast-Enhanced Cardiac MRI

To accurately characterize disease progression and quantify pathophysiological remodeling in the heart, clinical cardiology employs cardiac magnetic resonance (CMR) imaging with contrast enhancement (LGE) to visualize scarring and fibrosis in the ventricles. However, LGE-CMR image analysis is a labor-intensive process prone to large inter-observer variability that requires significant training and expertise. We developed a novel fully-automated anatomically-informed deep learning solution for LV and scar/fibrosis segmentation and clinical feature extraction from LGE-CMR.

‍

Publication on arXiv

Using Digital Hearts with Fat Infiltration to Predict Ventricular Tachycardia Ablation Targets

Patients suffering from ventricular tachycardia (VT) often undergo a clinical procedure called catheter ablation. Unfortunately, post-ablation VT recurrence is high due in part to a lack of precision. In this project, we developed a workflow that combines the fat infiltration distribution as identified on CT with computational heart modeling to predict optimal VT ablation targets. If deployed clinically, this technology has the potential to drastically improve ablation precision and improve patient outcomes. This work is published online in the journal Circulation: Arrhythmia & Electrophysiology (Personalized Digital-Heart Technology for Ventricular Tachycardia Ablation Targeting in Hearts With Infiltrating Adiposity | Circulation: Arrhythmia and Electrophysiology (ahajournals.org)).

Hemodynamic Analysis Predicts Stroke Risk in AF Patients Undergoing LAA Closure

AF patients are at fivefold higher risk for stroke than normal population. LAA closure procedure is most common to reduce risk of stroke for AF patients. We have developed a personalized blood-flow analysis based stroke risk predictor for AF patients undergoing LAA closure. In a proof-of-concept study on four AF patients undergoing LAA closure procedure, we show that for patients having stroke or transient ischemic attack (TIA) after the procedure, the LAA closure may not substantially reduce low-flow zone in the LA, predisposing them to risk of future stroke/TIA. The stroke prediction tool could potentially identify such cases, enabling a priori assessment of the efficacy of LAA closure devices for AF patients, as well as close monitoring of patients after LAA closure.

‍

Combining Machine Learning and Cardiac Modeling for Clinical Outcome Prediction

Machine learning and computational modeling have become increasingly pervasive in medicine. The goal of these projects is to combine the two into a clinical risk prediction algorithm, in what we call the Computational Heart and Artificial Intelligence (CHAI) approach. Preliminary work has focused on both atrial and ventricular modeling and demonstrates that the CHAI approach out-performs common clinical risk stratification methodologies.

  1. Prediction risk of atrial fibrillation recurrence following pulmonary vein isolation
  2. Ventricular arrhythmia risk prediction in patients with cardiac sarcoidosis using MRI-PET fusion mechanistic heart modeling and machine learning

‍

‍

‍

FDA-approved randomized clinical trial in patients with AF

In a proof-of-concept study, published in Nature Biomedical Engineering, we have demonstrated  the utility of personalized computational atrial modeling to guide ablation of persistent AF (psAF). This novel technology (termed OPTIMA, OPtimal Target Identification via Modeling of Arrhythmogenesis) is based on non-invasive patient-specific anatomic and tissue data from late gadolinium enhancement cardiac MRI (LGE-CMR) and simulation of cardiac electrical function to personalized ablation targets for psAF patients.

We now have FDA-approval for a 160-patient randomized clinical trial to demonstrate the utility of OPTIMA in patients with persistent atrial fibrillation and fibrosis. The approach is termed OPTIMA. The PIs are Drs. Trayanova, Calkins, and Spraag.

Identifying Ventricular Tachycardia Risk in Patients with Hypertrophic Cardiomyopathy

Current risk stratification criteria are inadequate in identifying patients with hypertrophic cardiomyopathy (HCM) at risk for ventricular tachycardia (VT) and in need of an implantable cardioverter defibrillator (ICD). Our personalized virtual heart technology incorporates a fusion of contrast-enhanced LGE-MRI and post-contrast T1 mapping to reconstruct the fibrotic remodeling that occurs in HCM. In silico rapid pacing is used to assess VT inducibility which may improve the identification of patients with HCM in need of an ICD.

Copyright © Natalia Trayanova 2016
What's NewOur PeopleGallery/ProjectsNews ArchivePublications