Atrial fibrillation driver identification through regional mutual information networks: a modeling perspective

Figures

• 

  Fig. 1. Clinical regional high-resolution local activation time mapping. a A schematic of high-resolution mapping catheters with flower-shaped open spline and grid-type parallel spline configurations. b A schematic of a catheter spline representing mapping electrodes and the spacing between them. c A schematic representing regional local activation time maps obtained throughout the atrium with a high-resolution flower-shaped open spline catheter. The colored region represents an individual map

• 

  Fig. 2. Regional mutual information graph approach. A) A schematic representing two mapping electrodes positioned on cardiac tissue. The action potentials at the electrode locations are used to estimate the mutual information between the recording sites. B) A graph with each node representing a recording electrode, with electrodes from panel A highlighted. The connections between nodes (edges) represent mutual information values above a set threshold. C) A schematic illustrating subgraph analysis. A set of nodes connected to a single central node thereby forming a subgraph is highlighted in dark grey. Edges highlighted in red represent the shortest paths between subgraph nodes that do not connect to the central node. The average of these path lengths over multiple subgraphs represents local efficiency. D) A series of electrode positions shown in grey overlayed over a series of simulated voltage maps demonstrate data extraction using a particular electrode geometry. E) A mutual information matrix demonstrating the mutual information between each pair of electrodes. Increased mutual information is shown in yellow. F) A binary adjacency matrix calculated by setting a threshold for mutual…

• 

  Fig. 3. Local efficiency signature of rotational activation. a Schematic representation of variable catheter positioning. Red crosses represent the center of the electrode array within a single time point of simulated rotational activation in the 3D model. b Local efficiency as a function of information threshold plotted in black for one of six grid type catheter positions within simulated tissue (3D model rotational activation). Corresponding third degree polynomial fit is shown in red. c First derivative of local efficiency for 3D tissue simulations with irregular (black) and rotational (red) activation patterns. The shaded region represents the 95% confidence interval based on varied grid type catheter position within the simulated tissue (n = 6). The region between the 40 and 80% information threshold used in further analysis is highlighted in blue. d Comparison of the average first derivative in the 40–80% information threshold range for irregular (black) and rotational (red) activation patterns in 2D and 3D simulations. Error bars represent the 95% confidence interval

• 

  Fig. 4. Local efficiency for remodeling classification. A) Representative local efficiency calculated for graphs with increasing information thresholds based on simulations that incorporate 10% remolding and a grid type catheter with 2 mm interelectrode spacing. B) Representative local efficiency calculated for graphs with increasing information thresholds based on simulations that incorporate 50% remolding and a grid-type catheter with 2 mm interelectrode spacing. C) Representative local efficiency calculated for graphs with increasing information thresholds based on simulations that incorporate 90% remolding and a grid catheter with 2 mm interelectrode spacing. D) The rate of change in local efficiency for 40–80% information thresholds for grid-type catheters and flower-shaped open spline-type catheters. Derivatives are averaged for 90% (red), 50% (orange), and 10% (black) remodeling simulations based on 6 catheter positions within the simulated tissue. Error bars represent the 95% confidence interval

• 

  Fig. 5. Demonstration of a local efficiency curve simulating from a clinical AF map (Subject 841–004, Sarkozy et al. 2020 [25]), which could be integrated into CARTOFINDER rotational activity analysis interface in the Carto 3 mapping system. The red rectangle within the LAT map highlights the anterior wall region of interest analyzed. The putative insert displays local efficiency values and the corresponding polynomial fit as a function of information threshold