Comparison of dipole-based and potential-based ECGI methods for premature ventricular contraction beat localization with clinical data

Figures

• 

  FIGURE 1. (A, B) Top row showing the anterior and posterior views of the body with the electrodes attached and the mesh generated from the body with the electrode positions marked on the torso surface, respectively. Four different heart–torso geometries used in this study are given in (C) of the bottom row. Homogeneous models only include the heart and torso surfaces (Epi–HT and EpiEndo–HT). Lungs are included in the Epi inhomogeneous model (Epi–HLT), and both blood and the lungs are included in the inhomogeneous EpiEndo model (EpiEndo–HBLT).

• 

  FIGURE 2. Summary of the ECGI pipelines used in this study. EP-Solutions data set geometry is only available in terms of the EpiEndo heart surface with the homogeneous torso. The Bratislava data set has both the Epi and EpiEndo surfaces defined, as well as inhomogeneities (lungs and intracavitary blood), as described in Section 2.4.

• 

  FIGURE 3. Boxplot representations of the LE values obtained for both data sets, both source models [dipole-based (A) and potential-based (B)], and all heart–torso models, for all patients. The red line in the boxplots corresponds to the median LE value for all patients, and the upper and lower boundaries of the boxes correspond to the third (Q3) and the first (Q1) quartiles, with IQR = Q3-Q1.

• 

  FIGURE 4. Ground truth pacing locations (black marks: circle—LV; diamond—RV) and the estimated pacing locations for the potential-based (squares) and the dipole-based (triangles) source models on the EpiEndo geometries for the EP-Solutions data set. Different colors are used to denote LV and RV pacing site estimates (blue—RV and magenta—LV). Top row: base to apex view and bottom row: posterior view of the heart. The left and right ventricles for the geometries in both rows are labeled as LV and RV, respectively.

• 

  FIGURE 5. Ablation points (black circles) and the estimated pacing locations for the potential-based (squares) and the dipole-based (triangles) source models on the EpiEndo geometries for the Bratislava data set. Different colors are used to denote different heart–torso geometries used in this study. Top row: base to apex view and bottom row: posterior view of the heart. The left and right ventricles for the geometries in both rows are labeled as LV and RV, respectively.

• 

  FIGURE 6. AT maps corresponding to the potential-based solutions for P3 (A) and P8 (B) of the Bratislava data set, all with the same view angle, including the estimated pacing location (white circles) and the ground truth PVC origin (black circles). Note that the ground truth PVC origins are marked on the EpiEndo surface; thus, the corresponding marking is not always visible on the Epi surface if the ablation point is on the endocardium.

• 

  FIGURE 7. RRE maps corresponding to the dipole-based solutions for P3 (A) and P6 (B) of the Bratislava data set, all with the same view angle, including the estimated pacing location (white circles) and the ground truth PVC origin (black circles). The PVC origin markings are similar to those in Figure 6.