A Comprehensive Evaluation of Cardiac Electro-Anatomic Mapping Using Body Surface Potentials

Abstract

The ongoing development of inverse techniques for mapping cardiac electrical activity from body surface potentials is driven by the general expectation that they can provide clinically useful information about heart rhythm disturbances. To date, a complete in-vivo validation of this inverse mapping approach has not been performed in large animals or humans. The principal objectives of this thesis were to evaluate the accuracy and reliability of inverse algorithms to reproduce cardiac electrical maps from high-density body surface recordings, and to explore whether they could complement established endocardial mapping techniques to aid catheter ablation therapies. A novel experimental pig model was developed to address these issues. High-density electropotential signals were mapped simultaneously from precisely determined 3D locations on the body and epicardial surface of the ventricles during sinus rhythm, ventricular pacing and VT. Concurrent electrical activity from the LV endocardium was reconstructed using a multi-electrode balloon catheter. Finally, torso anatomies were acquired post-mortem with MRI and used to develop experiment-specific volume conductor models; a homogeneous model with isotropic electrical properties, and an inhomogeneous model including subcutaneous fat, lungs and a simplified anisotropic skeletal muscle layer. Body surface potentials were simulated from epicardial recordings using homogeneous and inhomogeneous torso models. Results were compared with directly measured body surface potentials. Results demonstrated the potential difference between extrema predicted by a homogeneous forward model were 55-68% greater than observed, and the attenuation of potentials adjacent to extrema were also 21-130% greater (p-value < 0.05). Peak potentials were also inaccurately localised, with the Euclidean distance and orientation of the axis between them significantly different (p-values < 0.05). Inclusion of inhomogeneous electrical properties in the forward model not only reduced the potential magnitudes (14-19% smaller than predicted with a homogeneous model), but also improved the spatial precision of body surface potential maps, with significant improvements to the potential attenuation, and the localisation of potential extrema (p-values < 0.05). Despite these improvements, significant disparity remained between the forward simulated body surface potentials and those recorded (p-values < 0.05). The validity of the inverse model, using Greensite’s spatial and temporal regularisation method, was assessed by comparing measured and reconstructed epicardial potentials through potential maps, electrograms and activation patterns. Results demonstrated that, using a homogeneous volume conductor model, the general topology of simple epicardial ventricular excitation profiles were accurately reconstructed, with CCs of 0.76 0.10 and rRMSEs of 0.21 0.11 between recorded and reconstructed activation times. Despite this, epicardial sources were not localised with high precision (19 12 mm), and fine spatial and temporal details were not captured, including low voltage deviations in electrograms. The effect of inhomogeneous electrical properties on inverse solutions was also determined, and while the magnitude of inverse epicardial potentials more closely reflected those directly measured (p-values < 0.001), no other qualitative or quantitative improvement was seen to these inverse solutions when compared to a uniform isotropic model (p-values < 0.05). The inclusion of an ischaemia model allowed the validation of inverse methods using more complex cardiac rhythms, and their clinical utility could be further assessed. Noninvasively reconstructed epicardial potentials from body surface signals predicted and localised electrically altered myocardium resulting from acute ischaemia. These regions could be defined both by large ST-segment elevation and conduction delay over the ischaemic tissue. However, the inverse model was unable to resolve fine spatial or temporal details of ischaemia in epicardial potentials or activation maps. Despite this, the general pattern of macro-reentrant circuits were qualitatively resolved and the location of epicardial exit sites consistently identified to within 20 mm. Using simultaneously acquired endocardial potentials, the underlying mechanism associated with these forms of VT could also be identified. These results suggested that inverse mapping can play a clinically useful role. Inverse mapping has been shown to identify the epicardial mechanisms underlying a reentrant VT. Although precise localisation was not possible, the general location of an epicardial exit site was resolved. This information is clinically relevant as current ablation procedures are often hampered by long procedure times, with the majority of time dedicated to localising the site of ablation. Having the information on the nature of the arrhythmia gives guidance to these procedures in terms of not only a starting point to identifying ablation sites. If these techniques can be developed to give real-time information this would further aid clinical ablation procedures as the driving surface could be detected much quicker.