Left atrial imaging and registration of fibrosis with conduction voltages using LGE-MRI and electroanatomical mapping
Introduction
Atrial Fibrillation (AF) is a heart rhythm disorder of the left atrium (LA) that is associated with an increased risk of stroke and death. A 2010 study estimated that 2.7 to 6.1 million people in the United States suffer from AF, and this number is expected to increase to 12.1 million in 2030 [1]. AF can be classified as paroxysmal, persistent, and longstanding persistent (also known as ‘permanent’ AF) [2]. While catheter ablation is highly successful for most patients with paroxysmal AF, it is much less effective for patients with persistent and longstanding persistent AF [3]. Despite advancements in catheter ablation technology, its success rate remains as low as 45–50% after treatment [4]. These considerable treatment challenges have motivated a number of experimental studies, including some that explored the correlation between the extent of fibrillation burden and the degree of fibrosis in the LA [5,6]. Importantly, AF has been associated with structural changes to the extracellular matrix of the myocardium, including collagen deposition—a process known as fibrosis. For example, Verma [7] showed that patients with evidence of fibrosis in the LA at the time of ablation resulted in nearly half the success rate of those without (43 vs. 81%).
Electroanatomical voltage mapping systems (EAMs) can identify the presence, location, and extent of fibrosis in the LA by detection of abnormally conductive or ‘low voltage’ tissue regions [8]. These systems enable operators to position the catheters without fluoroscopic guidance [9]. Areas of abnormal myocardium (‘low bipolar voltage regions’) from fibrotic tissue in the LA are considered as potential targets for ablation [10]. Catheter ablation is an invasive procedure in which radio frequency (RF) energy is delivered through a catheter to deliberately destroy the small regions of heart tissue that are responsible for creating abnormal electrical signals. For example, pulmonary vein (PV) isolation is a common procedure, where the tissue bordering the pulmonary veins (PVs) is targeted for ablation in order to prevent abnormal wave propagation to the rest of myocardium. When electroanatomical voltage mapping (EAM) data is fused with other imaging sources, such as magnetic resonance (MR) Imaging or Computed Tomography (CT), it may provide crucial information for guiding catheter ablation by locating fibrosis.
Together with EAMs, pre-procedural MR can provide a detailed depiction of the LA and the PVs [11]. Late gadolinium enhancement (LGE)-MR is also playing an increasingly important role in the depiction of myocardial fibrosis [12] and can provide potentially relevant information prior to ablation. In late gadolinium enhancement-magnetic resonance imaging (LGE-MRI), images are acquired 10–20 min following a bolus injection of a gadolinium (Gd) based contrast agent, which allows the contrast to distribute to the extravascular-extracellular spaces, and accumulate, in regions of fibrosis. As the use of LGE-MRI increases in pre-ablation planning, there is an urgent need for registration techniques that will enable us to compare the spatial extent of low bipolar voltage areas on EAMs with hyper-enhanced regions on late gadolinium enhancement-magnetic resonance (LGE-MR) images. Several studies have demonstrated that fibrosis regions of the LA detected using LGE-MRI prior to ablation can assist in predicting ablation outcome [[13], [14], [15]]. Leonardi [14] reported that an accurate morphological three-dimensional (3D) left atrial model produced by magnetic resonance imaging (MRI) can improve the cardiac ablation strategy. Romero [15] described the clinical impact of incorporating LGE-MRI data to EAMs, specifically that regions of LGE-MR enhancement may assist in visualizing the fibrosis border zone in the LA.
Although others have reported significant correlations between EAMs and LGE-MRI with respect to the total burden of fibrosis in the LA [16,17], the spatial correlation between regions of low bipolar voltage on EAMs and regions of hyperenhancement on LGE-MRI depends on the accurate alignment of data acquired by these two disparate image modalities (i.e., registration). To our knowledge, registration methods for integration of LGE-MRI and EAMs of the LA have not been adequately described in the literature [18,19]. Therefore, we propose a computational registration and evaluation strategy, which was validated using clinical data acquired from a prospective study of patients with persistent AF.
The primary objective of this work is to develop and validate a registration method for aligning the left atrial surface determined from LGE-MRI to the left atrial surface depicted by EAMs. After implementation of this registration strategy, the secondary objective is to evaluate the spatial correlation between fibrosis depicted in LGE-MRI determined by several segmentation algorithms and regions of low bipolar voltage on EAMs.
Section snippets
Study subjects and image acquisition
Twenty patients (mean age: 64 years, age range: 45–76 years) participated in this prospective study of persistent AF. The study was approved by the Institutional Human Research Ethics Board, and a written consent was obtained from all participants. 3D LGE-MR images were acquired prior to first-time ablation (median duration: 57 days), and EAMs data were acquired immediately prior to ablation.
All patients underwent MRI examination on a 1.5 Tesla clinical scanner (Siemens Aera, Siemens Medical
Overview of our approach
In image processing, registration involves aligning two or more objects into the same spatial scene [24]. In this work, we used a point cloud to represent the left atrial geometry, as isolated from LGE-MR images and EAMs data. In this context, a point cloud is a collection of coordinate points, XY(Z), in space, which can be converted to create a surface of the LA using Delaunay triangulation.
Our proposed registration pipeline is shown in Fig. 2, which consists of manual, affine, and non-rigid
Segmentation of fibrosis in LGE-MRI
In this study, we applied the image intensity normalization method advanced by Giannakidis [33]. We normalized the image intensity of the voxels within the left atrial wall contours by that of the left atrial blood-pool [33] using the following;where is normalized left atrial wall intensity, is the mean value of the left atrial blood-pool signal I, is the standard deviation value of left atrial blood-pool signal I, and I(h) is left atrial wall intensity for
Spatial correlation of fibrosis determined by LGE-MRI to low bipolar voltage regions on EAMs
We projected vertices labeled as fibrosis in the left atrial wall determined from LGE-MR, created using intensity-threshold-based techniques, on EAMs to evaluate their spatial correlation. We used k-d tree method [34] based on the Euclidean distance (ED) metric to locate the closest point on the target surface. In the Euclidean space, the minimum distance between two points is drawing a straight line. During k-d tree search, each vertex of fibrosis in the left atrial wall was regarded as a
Comparison of our method to a state-of-the-art technique
We compared results of our method to the coherent point drifts (CPD) [38], which is an alternative to the NICP algorithm. We selected the CPD algorithm inspired by Liang [39] as a comparison to our NICP algorithm. The CPD algorithm considers the alignment of target and source datasets as a probability density estimation problem and coherently re-parameterizes Gaussian mixture model centroid locations with rigid parameters using maximization step of the Expectation-maximization algorithm. The
Evaluation of our registration approach
We used a landmark-based evaluation method by measuring the distance between corresponding landmark points on the two surfaces in ParaView. Fig. 4 shows the location of the chosen landmarks. To choose distinctive features in the LA, two vertices in the diagonal direction at each PV were selected under careful visual inspection. For left PVs, in the middle of a short vestibule or funnel-like common vein between LSPV and LIPV, the adjacent edges of the posterior and anterior wall were chosen as
Results
We summarized the results of our landmark-based TRE for our method and the alternative method (CPD) in Table 1. Manual alignment reduced the TRE from 368.94 mm to 25.06 mm, which was decreased further to 20.97 mm after affine ICP (Wilcoxon paired test between manual alignment and affine ICP, p < 0.05). NICP improved the TRE to 0.61 mm whereas CPD only improved the TRE to 1.8 mm (Wilcoxon paired test between NICP and CPD, p < 0.01). The TRE reduction associated with the NICP step was
Discussion
The objective of this study was to develop and evaluate a method for registering the left atrial surface from EAMs to the one determined from LGE-MR images. Such a method would enable the merging of complementing information from both LGE-MRI with EAMs, and potentially assist with ablation strategies for patients with AF. We described a registration pipeline based on the ICP algorithm, which was evaluated by measuring distance between manually picked anatomical landmarks in left atrial surface
Conclusions
We developed and validated a method for registering the left atrial surface on LGE-MRI to one on EAMs based on three integration steps. The results demonstrated that the integration of left atrial point cloud outperformed the comparative method verified by our evaluation methods. Furthermore, a substantial correlation between low bipolar voltage regions on EAMs and hyper-enhanced regions on LGE-MRI was demonstrated. The development of registration and evaluation methods can be utilized to
Disclosure of conflicts of interest
The authors have no relevant conflicts of interest to disclose.
Acknowledgements
Funding for this study was supported by Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery Grant and the Cardiac Arrhythmia Network of Canada (CANet) Strategic Research Grant program (SRG-15-P06-001) as part of the Networks of Centres of Excellence of Canada. The author is a member of the Cardiac Arrhythmia Network of Canada (CANet) HQP Association for Trainees (CHAT). The authors gratefully acknowledge the technical contributions of Sophie Chenier, MRT, and Owen
References (46)
- et al.
Structural abnormalities in atrial walls are associated with presence and persistency of atrial fibrillation but not with age
J. Am. Coll. Cardiol.
(2011) Pre-existent left atrial scarring in patients undergoing pulmonary vein antrum isolation: an independent predictor of procedural failure
J. Am. Coll. Cardiol.
(2005)A new approach for catheter ablation of atrial fibrillation: mapping of the electrophysiologic substrate
J. Am. Coll. Cardiol.
(2004)3D slicer as an image computing platform for the quantitative imaging Network
Magn. Reson. Imaging
(2012)- et al.
Object modeling by registration of multiple range images
- et al.
Nonrigid iterative closest points for registration of 3D biomedical surfaces
Optic Laser. Eng.
(2018) The association of left atrial low-voltage regions on electroanatomic mapping with low attenuation regions on cardiac computed tomography perfusion imaging in patients with atrial fibrillation
Heart Rhythm
(2015)Heart disease and stroke Statistics'2017 update
(2017)Acc/aha/esc 2006 guidelines for the management of patients with atrial fibrillation: a report of the American college of cardiology/American heart association task force on practice guidelines and the European society of cardiology committee for practice
Circulation
(2006)HRS/EHRA/ECAS EXPERT CONSENSUS STATEMENT 2012 HRS/EHRA/ECAS expert consensus statement on catheter and surgical ablation of atrial fibrillation: recommendations for patient selection, procedural techniques, patient management and follow-up, Definitions
E. Eur Niv Ad.
(2012)
Approaches to catheter ablation for persistent atrial fibrillation
N. Engl. J. Med.
The impact of weeks of atrial fibrillation on left atrial and ventricular structure and function
J. Thorac. Cardiovasc. Surg.
Current Status and Future of Cardiac Mapping in Atrial Fibrillation
Principles of electroanatomic mapping
Indian Pacing Electrophysiol. J.
Computational shape models characterize shape change of the left atrium in atrial fibrillation
Clin. Med. Insights Cardiol.
Improved multimodality data fusion of late gadolinium enhancement MRI to left ventricular voltage maps in ventricular tachycardia ablation
IEEE Trans. Biomed. Eng.
Advances in imaging for atrial fibrillation ablation
Radiol. Res. Pract.
Volumetric identification of left atrial fibrosis from delayed enhancement magnetic resonance imaging in atrial fibrillation: preliminary results
Comput. Cardiol.
CMR imaging for the evaluation of myocardial stunning after acute myocardial infarction: a meta-analysis of prospective trials
Eur. Heart J. Cardiovasc. Imag.
Detection and quantification of left atrial structural remodeling with delayed-enhancement magnetic resonance imaging in patients with atrial fibrillation
Circulation
Evaluation of left atrial lesions after initial and repeat atrial fibrillation ablation; Lessons learned from delayed-enhancement MRI in repeat ablation procedures
Circ. Arrhythmia Electrophysiol.
Integration of 3D electroanatomic maps and magnetic resonance scar characterization into the navigation system to guide ventricular tachycardia ablation
Circ. Arrhythmia Electrophysiol.
Head-to-head comparison of contrast-enhanced magnetic resonance imaging and electroanatomical voltage mapping to assess post-infarct scar characteristics in patients with ventricular tachycardias: real-time image integration and reversed registration
Eur. Heart J.
Cited by (5)
Atrial location optimization by electrical measures for Electrocardiographic Imaging
2020, Computers in Biology and MedicineCitation Excerpt :For the entire database, composed by the 28 different simulations in which the optimization algorithm was run twice (56 tests), the average error between the original and the final position is shown in Fig. 2, for an initial error of 4.3 cm and 52° in random directions. The final location for the 56 test was obtained after 17 [9–33] iterations in 748 [434–1686] seconds. For displacements (Panel A), the location algorithm provided a deviation between the actual and the final location of 0.2 ± 0.2 cm in the X axis, 0.2 ± 0.3 cm for the Y axis and 0.4 ± 0.3 cm for the Z axis, resulting in an absolute error of 0.5 ± 0.5 cm in terms of Euclidean distance.
Image-to-Patient Registration in Computer-Assisted Surgery of Head and Neck: State-of-the-Art, Perspectives, and Challenges
2023, Journal of Clinical MedicineA systematic review and meta-analysis of the safety and efficacy of left atrial substrate modification in atrial fibrillation patients with low voltage areas
2022, Frontiers in Cardiovascular MedicineCorrelation between Cardiac MRI and Voltage Mapping in Evaluating Atrial Fibrosis: A Systematic Review
2022, Radiology: Cardiothoracic Imaging