Providing therapies tailored to each patient is the vision of precision medicine, enabled by the increasing ability to capture extensive data about individual patients. In this position paper, we argue that the second enabling pillar towards this vision is the increasing power of computers and algorithms to learn, reason, and build the ‘digital twin’ of a patient. Computational models are boosting the capacity to draw diagnosis and prognosis, and future treatments will be tailored not only to current health status and data, but also to an accurate projection of the pathways to restore health by model predictions. The early steps of the digital twin in the area of cardiovascular medicine are reviewed in this article, together with a discussion of the challenges and opportunities ahead. We emphasize the synergies between mechanistic and statistical models in accelerating cardiovascular research and enabling the vision of precision medicine.

The human heart beats as a result of multiscale nonlinear dynamics coupling subcellular to whole organ processes, achieving electrophysiologically-driven mechanical contraction. Computational cardiac modelling and simulation have achieved a great degree of maturity, both in terms of mathematical models of underlying biophysical processes and the development of simulation software.

Acute ischemia is a major cause of sudden arrhythmic death, further promoted by potassium current blockers. Macro-reentry around the ischemic region and early afterdepolarizations (EADs) caused by electrotonic current have been suggested as potential mechanisms in animal and isolated cell studies. However, ventricular and human-specific arrhythmia mechanisms and their modulation by repolarization reserve remain unclear. The goal of this paper is to unravel multiscale mechanisms underlying the modulation of arrhythmic risk by potassium current (IKr) block in human ventricles with acute regional ischemia. A human ventricular biophysically-detailed model, with acute regional ischemia is constructed by integrating experimental knowledge on the electrophysiological ionic alterations caused by coronary occlusion. Arrhythmic risk is evaluated by determining the vulnerable window (VW) for reentry following ectopy at the ischemic border zone. Macro-reentry around the ischemic region is the main reentrant mechanism in the ischemic human ventricle with increased repolarization reserve due to the ATP-sensitive potassium current (IK(ATP)) activation. Prolongation of refractoriness by 4% caused by 30% IKr reduction counteracts the establishment of macro-reentry and reduces the VW for reentry (by 23.5%). However, a further decrease in repolarization reserve (50% IKr reduction) is less anti-arrhythmic despite further prolongation of refractoriness. This is due to the establishment of transmural reentry enabled by electrotonically-triggered EADs in the ischemic border zone. EADs are produced by L-type calcium current (ICaL) reactivation due to prolonged low amplitude electrotonic current injected during the repolarization phase. Electrotonically-triggered EADs are identified as a potential mechanism facilitating intramural reentry in a regionally-ischemic human ventricles model with reduced repolarization reserve.

The left atrium (LA) can change in size and shape due to atrial fibrillation (AF)-induced remodeling. These alterations can be linked to poorer outcomes of AF ablation. In this study, we propose a novel comprehensive computational analysis of LA anatomy to identify what features of LA shape can optimally predict post-ablation AF recurrence. To this end, we construct smooth 3D geometrical models from the segmentation of the LA blood pool captured in pre-procedural MR images. We first apply this methodology to characterize the LA anatomy of 144 AF patients and build a statistical shape model that includes the most salient variations in shape across this cohort. We then perform a discriminant analysis to optimally distinguish between recurrent and non-recurrent patients. From this analysis, we propose a new shape metric called vertical asymmetry, which measures the imbalance of size along the anterior to posterior direction between the superior and inferior left atrial hemispheres. Vertical asymmetry was found, in combination with LA sphericity, to be the best predictor of post-ablation recurrence at both 12 and 24 months (area under the ROC curve: 0.71 and 0.68, respectively) outperforming other shape markers and any of their combinations. We also found that model-derived shape metrics, such as the anterior-posterior radius, were better predictors than equivalent metrics taken directly from MRI or echocardiography, suggesting that the proposed approach leads to a reduction of the impact of data artefacts and noise. This novel methodology contributes to an improved characterization of LA organ remodeling and the reported findings have the potential to improve patient selection and risk stratification for catheter ablations in AF.

Background: The accuracy of ECG imaging (ECGI) in structural heart disease remains uncertain. This study aimed to provide a detailed comparison of ECGI and contact-mapping system (CARTO) electrograms. Methods: Simultaneous epicardial mapping using CARTO (Biosense-Webster, CA) and ECGI (CardioInsight) in 8 patients was performed to compare electrogram morphology, activation time (AT), and repolarization time (RT). Agreement between AT and RT from CARTO and ECGI was assessed using Pearson correlation coefficient, ρ _AT and ρ _RT , root mean square error, E _AT and E _RT , and Bland-Altman plots. Results: After geometric coregistration, 711 (439–905; median, first-third quartiles) ECGI and CARTO points were paired per patient. AT maps showed ρ _AT =0.66 (0.53–0.73) and E _AT =24 (21–32) ms, RT maps showed ρ _RT =0.55 (0.41–0.71) and E _RT =51 (38–70) ms. The median correlation coefficient measuring the morphological similarity between the unipolar electrograms was equal to 0.71 (0.65–0.74) for the entire signal, 0.67 (0.59–0.76) for QRS complexes, and 0.57 (0.35–0.76) for T waves. Local activation map correlation, ρ _AT , was lower when default filters were used (0.60 (0.30–0.71), P =0.053). Small misalignment of the ECGI and CARTO geometries (below ±4 mm and ±4°) could introduce variations in the median ρ _AT up to ±25%. Minimum distance between epicardial pacing sites and the region of earliest activation in ECGI was 13.2 (0.0–28.3) mm from 25 pacing sites with stimulation to QRS interval <40 ms. Conclusions: This simultaneous assessment demonstrates that ECGI maps activation and repolarization parameters with moderate accuracy. ECGI and contact electrogram correlation is sensitive to electrode apposition and geometric alignment. Further technological developments may improve spatial resolution.

To identify key structural and electrophysiological features explaining distinct electrocardiogram (ECG) phenotypes in hypertrophic cardiomyopathy (HCM). Human heart–torso anatomical models were constructed from cardiac magnetic resonance (CMR) images of HCM patients, representative of ECG phenotypes identified previously. High performance computing simulations using bidomain models were conducted to dissect key features explaining the ECG phenotypes with increased HCM Risk-SCD scores, namely Group 1A, characterized by normal QRS but inverted T waves laterally and coexistence of apical and septal hypertrophy; and Group 3 with marked QRS abnormalities (deep and wide S waves laterally) and septal hypertrophy. Hypertrophic cardiomyopathy abnormalities characterized from CMR, such as hypertrophy, tissue microstructure alterations, abnormal conduction system, and ionic remodelling, were selectively included to assess their influence on ECG morphology. Electrocardiogram abnormalities could not be explained by increased wall thickness nor by local conduction abnormalities associated with fibre disarray or fibrosis. Inverted T wave with normal QRS (Group 1A) was obtained with increased apico-basal repolarization gradient caused by ionic remodelling in septum and apex. Lateral QRS abnormalities (Group 3) were only recovered with abnormal Purkinje-myocardium coupling. Two ECG-based HCM phenotypes are explained by distinct mechanisms: ionic remodelling and action potential prolongation in hypertrophied apical and septal areas lead to T wave inversion with normal QRS complexes, whereas abnormal Purkinje-myocardial coupling causes abnormal QRS morphology in V4–V6. These findings have potential implications for patients’ management as they point towards different arrhythmia mechanisms in different phenotypes.