Inside-Out against Outside-In Inverse Solutions for Estimating Atrial Fibrillation Sources

Atrial fibrillation (AF) is the most frequently occurring cardiac arrhythmia. Further understanding of the drivers and maintainers of AF is still necessary to improve treatment strategies and outcome. This thesis focussed on identifying AF sources using body-surface mapping and non-contact endocardial mapping (NCM, inside-out inverse solution) data. Analysis was focussed on dominant frequency (DF) and highest dominant frequency (HDF) behaviour. Atrial sources on the atrial muscle were estimated from body surface mapping (BSM) using a heterogeneous torso model (outside-in inverse solution). The current work has shown that a significant difference exists between the DF and HDF behaviour on the BSM and NCM data. This shows the effect of the torso volume conductor on BSM data as previously suggested in simulation studies. This effect should be taken into account when analysing BSM data for diagnosing AF. In this work, a new inside-out inverse solution was developed based on equivalent double layer cardiac sources. This was compared to commercially available software (EnSite, St Jude Medical), which was considered the gold standard. This system is based on a potential-based transfer, and therefore only provides a “closer look” at the pericardial potentials, but does not estimate sources. Due to this variation in source model, significant differences could be found between the inverse solutions, leading to high regularisation parameters to force the home-made inverse solution to be as equal as possible to the gold standard. This also meant that high-frequency components could not be reproduced accurately with the home-made algorithm, leading to significant variation in DF and HDF behaviour between both inside-out inverse solutions. Lastly a double-layer based outside-in inverse solution was developed and compared to the gold standard inside-out inverse solution. Although both solutions originate from a different source model, it was possible to reconstruct simple AF behaviour based on the HDF behaviour.