Wavefront directionality and decremental stimuli synergistically improve identification of ventricular tachycardia substrate: insights from personalized computational heart models

Figures

• 

  Figure 1. Study workflow. In personalized 3D computational heart models, pacing sites were selected throughout the left ventricle on both the endocardium and epicardium. Three pacing protocols were delivered: (i) an initial S1, (ii) S2 delivered 360 ms after S1, and (iii) S2 delivered 310 ms after S1. Unipolar signal electrograms were simulated along the myocardial surface and substrate EAMs were constructed. From the unipolar signals, six EAM features were extracted: voltage amplitude, conduction velocity, isochronal crowding, fractionation index, electrogram duration, and frequency power. The distributions of these EAM features were then compared with the infarct remodelling distribution and the locations of critical VT sites. EAM, electroanatomic mapping; VT, ventricular tachycardia.

• 

  Figure 2. Concordance between S1 EAM features from different paced maps. (A) Relationship between percentage remodelling and EAM feature concordance. Colours represent the different EAM features; dashed line is the line of best fit. (B) Violin plots illustrating the EAM feature concordances in non-injured myocardium vs. infarct remodelling. (C) Example depicting EAM feature concordance in a heart with more remodelling vs. a heart with less remodelling. EAM, electroanatomic mapping.

• 

  Figure 3. Detection of infarct remodelling using S1 EAM features. F1 scores for logistic regression models predicting the presence of infarct remodelling using EAM features from (A) individual and (B) multiple pacing locations in hearts with more (left) vs. less remodelling (right). (C) Example of how well S1 EAM features from multiple maps identifies the infarct distribution in a heart with more vs. less remodelling. EAM, electroanatomic mapping.

• 

  Figure 4. Impact of decremental pacing on detection of infarct remodelling. F1 scores for logistic regression models predicting the presence of infarct remodelling using EAM features from both S1 and S2 maps paced from (A) individual and (B) multiple locations. Electroanatomic mapping features were ordered based on distance from infarct, starting with maps paced most proximally and ending with maps paced most distally (bottom to top). EAM, electroanatomic mapping.

• 

  Figure 5. Examples of VT circuits along with S1 EAM features. For each example shown, the top and bottom rows display the S1 EAM features from a proximally and distally paced map, respectively. White/light grey arrows indicate the re-entrant pathway of the VT circuit, connecting from the exit site to the end of the common pathway. The black/dark grey lines indicate the lines of functional conduction block. VT, ventricular tachycardia; EAM, electroanatomic mapping.

• 

  Figure 6. Identifying the inner VT circuit using EAM features. (A) Example of how outer (black) and inner (white) VT circuits are defined. Area under the receiver operating characteristics for logistic regression models identifying the inner VT circuit using EAM features from (B) a single pacing site and (C) multiple pacing sites. Top and bottom for (C) show the results from when EAM features from the 50 maps paced most proximally and 50 maps paced most distally to the VT circuit were used, respectively. VT, ventricular tachycardia; EAM, electroanatomic mapping.