Use of Digital Twins in the daily clinical cardiology practice: revolution or utopia?

David Filgueiras-Rama
Spanish National Center for Cardiovascular Research (CNIC)


Background: Catheter-based ablation has become the first line treatment for a number of cardiac arrhythmias, including infarct-related ventricular reentrant tachycardia (VT). However, current procedures often fail to eliminate VT isthmuses as a result of inability to accurately determine the arrhythmogenic substrate, especially in patients with hemodynamically untolerated or non-inducible VT. Virtual heart models have been proposed as a potential solution to identify ablation targets before the procedure and potentially decrease complication rates and procedure time.

Objective: Study ventricular models generated from animals and patients with infarct-related substrate and identify the impact of imaging resolution and different substrate parameters on 3D modelling characteristics and expected outcomes.

Methods and Results: We studied a group of 10 pigs with established myocardial infarction undergoing high-resolution electroanatomical mapping and in vivo and ex vivo cardiac magnetic resonance (CMR) imaging to generate animal specific heart models. Model-derived characteristics were compared among imaging techniques and correlation analyses were performed to determine the impact of the modelling substrate in experimental outcome prediction. Similar analysis was performed in 15 patients with chronic myocardial infarction undergoing CMR imaging characterization of the myocardial substrate using 2D and 3D sequences. Patient-specific models using the same electrophysiological parameters were generated to identify the impact of imaging resolution on virtual electrophysiological studies. The results, in humans and animals, show that the predictive performance of current individual-specific heart models is highly dependent on the modelling substrate generated from associated imaging techniques. Lower resolution images result in larger scar volumes (total scar, heterogeneous and dense scar) both in animals and patients. This substantially affects computational outcomes obtained on inducibility parameters and VT dynamics from the same patient. Conclusion: Individual-specific heart models are greatly affected by substrate characterization with current imaging techniques. Identification and validation of optimal substrate characterization is an essential step for future translational impact.