Preliminary findings on left atrial appendage occlusion simulations applying different endocardial devices.

Atrial fibrillation (AF) is one of the most investigated arrhythmias since it is associated with a five-fold increase in the risk of strokes. Left atrium dilation and unbalanced and irregular contraction caused by AF favour blood stasis and, consequently, stroke risk. The left atrial appendage (LAA) is the site of the highest clots formation, increasing the incidence of stroke in AF population. For many years oral anticoagulation therapy has been the most used AF treatment option available to decrease stroke risk. Unfortunately, several contraindications including bleeding risk increase, interference with other drugs and with multiorgan functioning, might outweigh its remarkable benefits on thromboembolic events. For these reasons, in recent years, other approaches have been designed, including LAA percutaneous closure. Unfortunately, nowadays, LAA occlusion (LAAO) is restricted to small subgroups of patients and require a certain level of expertise and training to successfully complete the procedure without complications. The most critical clinical problems associated with LAAO are represented by peri-device leaks and device related thrombus (DRT). The anatomical variability of the LAA plays a key role in the choice of the correct LAA occlusion device and in its correct positioning with respect to the LAA ostium during the implant. In this scenario, computational fluid dynamics (CFD) simulations could have a crucial role in improving LAAO intervention. The aim of this study was to simulate the fluid dynamics effects of LAAO in AF patients to predict hemodynamic changes due to the occlusion. LAAO was simulated by applying two different types of closure devices based on the plug and the pacifier principles on 3D LA anatomical models derived from real clinical data in five AF patients. CFD simulations were performed on the left atrium model before and after the LAAO intervention with each device. Blood velocity, particle washout and endothelial damage were computed to quantify flow pattern changes after the occlusion in relation to the thrombogenic risk. Our preliminary results confirmed an improved blood washout after the simulated implants and the capability of foreseeing thrombogenic risk based on endothelial damage and maximum blood velocities in different scenarios. This tool may help to identify effective device configurations in limiting stroke risk for patient-specific LA morphologies.

A global benchmark of algorithms for segmenting the left atrium from late gadolinium-enhanced cardiac magnetic resonance imaging.

Segmentation of medical images, particularly late gadolinium-enhanced magnetic resonance imaging (LGE-MRI) used for visualizing diseased atrial structures, is a crucial first step for ablation treatment of atrial fibrillation. However, direct segmentation of LGE-MRIs is challenging due to the varying intensities caused by contrast agents. Since most clinical studies have relied on manual, labor-intensive approaches, automatic methods are of high interest, particularly optimized machine learning approaches. To address this, we organized the 2018 Left Atrium Segmentation Challenge using 154 3D LGE-MRIs, currently the world's largest atrial LGE-MRI dataset, and associated labels of the left atrium segmented by three medical experts, ultimately attracting the participation of 27 international teams. In this paper, extensive analysis of the submitted algorithms using technical and biological metrics was performed by undergoing subgroup analysis and conducting hyper-parameter analysis, offering an overall picture of the major design choices of convolutional neural networks (CNNs) and practical considerations for achieving state-of-the-art left atrium segmentation. Results show that the top method achieved a Dice score of 93.2% and a mean surface to surface distance of 0.7 mm, significantly outperforming prior state-of-the-art. Particularly, our analysis demonstrated that double sequentially used CNNs, in which a first CNN is used for automatic region-of-interest localization and a subsequent CNN is used for refined regional segmentation, achieved superior results than traditional methods and machine learning approaches containing single CNNs. This large-scale benchmarking study makes a significant step towards much-improved segmentation methods for atrial LGE-MRIs, and will serve as an important benchmark for evaluating and comparing the future works in the field. Furthermore, the findings from this study can potentially be extended to other imaging datasets and modalities, having an impact on the wider medical imaging community.

A Proof of Concept for Computational Fluid Dynamic Analysis of the Left Atrium in Atrial Fibrillation on a Patient-Specific Basis.

Atrial fibrillation (AF) is associated with a fivefold increase in the risk of cerebrovascular events, being responsible of 15-18% of all strokes. The morphological and functional remodeling of the left atrium (LA) caused by AF favors blood stasis and, consequently, stroke risk. In this context, several clinical studies suggest that the stroke risk stratification could be improved by using hemodynamic information on the LA and the left atrial appendage (LAA). The goal of this study was to develop a personalized computational fluid dynamics (CFD) model of the LA which could clarify the hemodynamic implications of AF on a patient-specific basis. In this paper, we present the developed model and its application to two AF patients as a preliminary advancement toward an optimized stroke risk stratification pipeline.

Towards a repository of synthetic electrograms for atrial activation detection in atrial fibrillation.

BACKGROUND: Recently, the analysis of the spatio-temporal behavior of atrial fibrillation activation patterns has been widely investigated with the aim to better understand the arrhythmia implications on the heart electrical activity. Most of the proposed techniques are based on atrial activation timing detections. Unfortunately atrial activation timings are not easily recognizable on the electrograms (EGMs) and an approach to support the validation of such techniques is highly desirable. The aim of this study is to provide an effective workflow for the generation of synthetic unipolar atrial electrograms (SEGMs) in atrial fibrillation (AF) condition and with different levels of noise. METHOD: Real EGMs signals were obtained from a dataset of 6 subjects that underwent ablation. Each SEGM was obtained by modeling the three principal components of an EGM starting from real signals: atrial far-field (Afar), atrial near-field (Anear) and the ventricular far-field (Vfar). Afar was generated using an autoregressive model applied on segments from real EGMs not characterized by ventricular or atrial activations; Anear and Vfar were extracted directly from the real signals. A Gamma distribution and an atrio-ventricular node model were used to locate both Anear and Vfar on Afar, respectively. Three electrophysiologists with different levels of expertise evaluated the realism of the SEGMs on a set of 100 randomly selected signals including 50 EGMs and 50 SEGMs. Analysis was repeated by the three experts on a subset of 21 signals. RESULTS: The time required to generate the synthetic EGMs was less than 1 min once annotated EGMs are available. The cardiologists succeeded in distinguishing real from synthetic EGMs in 45%, 43% and 35% of the signals, respectively. By repeating the evaluation, 28%, 0% and 48% of signals were classified differently, including 67%, 52% and 36% of correct classifications. CONCLUSION: The proposed approach proved to be effective in producing SEGMs which are difficult to distinguish from real EGMs. This study provides a tool for realistic SEGM generation from real EGMs in AF condition with different levels of noise and at different AF rates. The tool may be easily adopted to obtain SEGMs in different arrhythmic conditions. SEGMs generated in this study are shared with the scientific community as a first step towards a repository of synthetic and real atrial signals supporting the benchmarking of new approaches to investigate AF.

The Impact of Left Atrium Appendage Morphology on Stroke Risk Assessment in Atrial Fibrillation: A Computational Fluid Dynamics Study.

Atrial fibrillation (AF) carries out a 5-fold increase in stroke risk, related to embolization of thrombi clotting in left atrium (LA). Left atrial appendage (LAA) is the site with the highest blood stasis which causes thrombus formation. About 90 % of the intracardiac thrombi in patients with cardioembolic events originally develop in the LAA. Recent studies have been focused on the association between LAA anatomical features and stroke risk and provided conflicting results. Haemodynamic and fluid dynamic information on the LA and mostly on the LAA may improve stroke risk stratification. Therefore, the aim of this study was the design and development of a workflow to quantitatively define the influence of the LAA morphology on LA hemodynamics. Five 3D LA anatomical models, obtained from real clinical data, which were clearly different as regard to LAA morphology were used. For each LAA we identified and computed several parameters describing its geometry. Then, one LA chamber model was chosen and a framework was developed to connect the different LAAs belonging to the other four patients to this model. These new anatomical models represented the computational domain for the computational fluid dynamics (CFD) study; simulations of the hemodynamics within the LA and LAA were performed in order to evaluate the interplay of the LAA shape on the blood flow characteristics in AF condition. CFD simulations were carried out for five cardiac cycles. Blood velocity, vorticity, LAA orifice velocity, residence time computed in the five models were compared and correlated with LAA morphologies. Results showed that not only complex morphologies were characterized by low velocities, low vorticity and consequently could carry a higher thrombogenic risk; even qualitatively simple morphologies showed a thrombogenic risk equal, or even higher, than more complex auricles. CFD results supported the hypothesis that LAA geometric characteristics plays a key-role in defining thromboembolic risk. LAA geometric parameters could be considered, coupled with the morphological characteristics, for a comprehensive evaluation of the regional blood stasis. The proposed procedure might address the development of a tool for patient-specific stroke risk assessment and preventive treatment in AF patients, relying on morpho-functional defintion of each LAA type.