Heart Digital Twins Predict Features of Invasive Reentrant Circuits and Ablation Lesions in Scar-Dependent Ventricular Tachycardia

Abstract

Background: Catheter ablation of scar-dependent ventricular tachycardia (VT) is frequently hampered by hemodynamic instability, long procedure duration, and high recurrence rates. Magnetic resonance imaging-based personalized heart digital twins may overcome these challenges by noninvasively predicting VT circuits and optimum ablation lesion sites. In this combined clinical and digital twin study, we investigated the relationship between digital twin-predicted VTs and optimum ablation lesion sets with their invasively mapped counterparts during clinical VT ablation.

Methods: A total of 18 patients with scar-dependent VT underwent digital twin creation based on preprocedural, contrast-enhanced cardiac magnetic resonance imaging. Using rapid pacing protocols, VT was simulated and ablation targets were derived that would terminate all possible VTs in the models. Patients subsequently underwent invasive VT ablation, including targeting of diastolic activity and optimum entrainment sites. Digital twin-predicted VT circuits and ablation lesions were compared with their invasive clinical counterparts.

Results: Forty-three clinical VTs and 92 digital twin VTs were induced. Diastolic activity was seen in 16 of 43 (37.2%) clinical VTs. Sensitivity, specificity, positive predictive, and negative predictive values for the detection of critical VT sites by digital twins were 81.3%, 83.8%, 21.7%, and 98.8%, respectively. At an American Heart Association-segment level, agreement between clinical VT critical sites and digital twin primary predicted sites was moderate, with a κ coefficient of 0.46 (±0.32; P≤0.001). Termination of VT with ablation was achieved at a digital twin-predicted site in 4 of 5 (80%) cases where attempted. A total of 426 of 709 (60.1%) lesions were within 5 mm of a predicted target site. In total, 54.0% (±28.9%) of the digital twin-predicted area was ablated per patient based on conventional mapping criteria.

Conclusions: Heart digital twin VT circuits and ablation targets accurately predict many features of their respective clinical counterparts but have some limitations in spatial resolution. Our findings demonstrate the significant potential of digital twin technology in guiding catheter ablation for scar-dependent VT.

Keywords: catheter ablation; heart rate; humans; magnetic resonance imaging; technology.

Figure 1.

Study overview. Personalized heart digital twin creation (top line): 3-dimensional late gadolinium-enhanced cardiac magnetic resonance (3D LGE-CMR) images are segmented and a personalized heart digital twin created. This model is tested for arrhythmia inducibility and optimum ablation targets to terminate all possible ventricular tachycardias (VTs) in the model. Clinical VT ablation (bottom line): Patients with scar-dependent VT undergo conventional electroanatomical mapping, which is merged with the digital twin. VT is induced and mapped for comparison with digital twin VT simulations. Final clinical ablation lesion set is compared with the digital twin-predicted ablation lesion set. EAM indicates electroanatomical map.

Figure 2.

Example of ventricular tachycardia (VT) isthmus mapped to inferior wall. A, VT activation map with putative channel (yellow borders) and mid-isthmus (green star). B, Substrate map shows inferior wall scar at this location. C, Predicted target sites overlayed onto VT activation map shows correlation with isthmus. D, Mid-diastolic potentials (yellow arrows) observed during VT at this area. E, 12 lead ECG of the VT. F, 12/12 pacemap match achieved by pacing at site of mid-diastolic potentials (MDPs).

Figure 3.

Representative induced VT from invasive mapping (top) matching closely with digital twin-modeled VT (bottom) utilizing an inferior wall isthmus in a patient with ischemic cardiomyopathy. Dotted arrows represent the wavefront activation pattern.

Figure 4.

Patient with nonischaemic cardiomyopathy where induced ventricular tachycardia (VT) isthmus was located at predicted target site. A, voltage map showed peri-aortic scar. B, VT activation map showed isthmus (yellow borders) at site of peri-aortic scar. C and D, Digital twin-predicted target sites (red region) overlaid onto substrate map and VT activation map with green star at isthmus. E, 12 lead ECG of induced VT shows atypical left-bundle block morphology, left-inferior axis VT, tachycardia cycle length=400 ms. F, Mid-diastolic potentials demonstrated at isthmus site (yellow arrows). G, entrainment with concealed fusion on 12-lead ECG. H, Entrainment response showing excellent postpacing interval of 398 ms.

Figure 5.

Example demonstrating close overlap of clinical ablation lesions with digital twin predictions. A, Final clinical ablation lesion set (demarcated in yellow) from an ischemic cardiomyopathy patient with septal scar. B, Digital twin predicted target sites overlayed on lesion set, showing close spatial relationship. C, Final clinical ablation lesion set for from an ischemic cardiomyopathy patient with inferior wall scar. D, Digital twin predicted target sites show close spatial relationship with final ablation set.