In this study, we consider Intra-Atrial Reentrant Tachycardia (IART), an arrhythmia in which a non-linear excitation wave is attaching to an anatomical surgical scar in the atrium. Since the underlying pathophysiology of IART significantly differs between individuals, patient-specific treatment is desired to reduce mortality rates. To guide this precision medicine, patient-specific computational models of cardiac electrophysiology, also called cardiac digital twins, can be created. However, integrating patient data into such computational models still comes with many challenges. State-of-the-art clinical imaging and electrophysiological mapping techniques come with measurement sparsity and uncertainty which limits the precision with which we can model cardiac tissue properties. Uncertainties need to be included into the model to assess the reliability of the digital twin when aiding clinical decisions.
The goal of this study is to develop a computational method that can enable patient-specific characterization of scared tissue, based on local catheter data and a phenomenological cardiac electrophysiological model. We account for limited precision due to noisy and sparse measurements by including uncertainties. To develop the method, we define a computational atrial model over a 2D-manifold. We propose a parameterization of the scar of moderate dimension by defining a vector that describes the location, orientation and shape of the scar. We develop a Bayesian formulation of the inference problem (defining an appropriate likelihood function for the data) and a Markov chain Monte Carlo based algorithm to sample the posterior distribution of these parameters based on local catheter data. The resulting posterior both provides an estimate of the scar parameters and quantifies the uncertainty due to the sparse and noisy measurements. Besides presenting methodological advances towards quantifying posterior uncertainty, we perform a study of how the sparsity and precision of the data influence posterior uncertainty on the scar parameters.