Patient-specific atrial digital twins calibrated to reproduce clinical activation patterns can partially capture activation-based EGM biomarker trends, but voltage-related behavior was not accurately reproduced, highlighting the need for improved modeling before simulated EGMs can fully reproduce clinical signal amplitudes.
Key Findings
Results
The multi-step calibration workflow significantly reduced discrepancies between simulated and clinical local activation time (LAT) maps, decreasing mean absolute error from 36.3 ms to 14.3 ms.
Calibration was performed on 20 patients with persistent atrial fibrillation (psAF).
Mean absolute error (MAE) was reduced from 36.3 ± 11.1 ms at baseline to 14.3 ± 3.8 ms after complete calibration (p < 0.01).
LAT maps were obtained after the third extrastimulus of the triple short-coupled extrastimuli pacing protocol (3-Extra).
The calibration included global and local diffusion adjustment and ionic channel remodeling optimization.
Results
Total depolarization time (TDT) errors were substantially reduced from 41.3 ms to 4.9 ms following calibration.
TDT errors decreased from 41.3 ± 21.3 ms at baseline to 4.9 ± 5.8 ms after complete calibration.
This reduction indicates substantial correction of baseline conduction discrepancies.
The study involved 20 psAF patients with anatomical meshes derived from CT scans.
Fibrotic regions were estimated from clinical bipolar EGM voltage maps.
Results
Simulated EGMs reproduced clinical trends in activation duration, EGM fractionation, and LAT variability between clinically annotated healthy and abnormal atrial tissue.
Differences between healthy and abnormal tissue biomarkers were statistically significant (p < 0.01).
The magnitude of differences in these biomarkers was smaller in simulations compared to clinical data.
Voltage-related behavior was not accurately reproduced in the simulated EGMs.
Biomarker trends were compared between clinically annotated healthy and abnormal atrial tissue regions.
Methods
Patient-specific atrial models were built using anatomical meshes derived from CT scans incorporating fiber direction, atlas-derived tissue heterogeneity, and fibrosis estimation from bipolar EGM voltage maps.
The cohort consisted of 20 patients with persistent atrial fibrillation (psAF).
Model inputs included fiber direction, atlas-derived tissue heterogeneity, and fibrotic regions estimated from clinical bipolar EGM voltage maps.
Electrophysiological properties were personalized through a multi-step calibration procedure.
Calibration steps included global and local diffusion adjustment and ionic channel remodeling optimization.
Discussion
Voltage-related EGM behavior was not accurately reproduced by the atrial digital twins, indicating a key limitation of the current modeling approach.
Despite successful reproduction of activation-based biomarker trends, voltage amplitudes in simulated EGMs did not match clinical signal amplitudes.
The authors identified the need for improved modeling of voltage-related mechanisms as a prerequisite before simulated EGMs can fully reproduce clinical signals.
This limitation was noted despite significant improvements in LAT and TDT metrics.
The authors stated this 'highlighting the need for improved modeling of voltage-related mechanisms before simulated EGMs can fully reproduce clinical signal amplitudes.'
Conclusions
The calibrated digital twins demonstrate feasibility for use as simulation tools to investigate atrial conduction abnormalities and potentially guide therapy-planning strategies.
Models were proposed as tools to investigate atrial conduction abnormalities.
Authors highlighted potential for guiding future therapy-planning strategies in persistent AF, for which current pharmacological and ablative therapies remain suboptimal.
The study involved 20 psAF patients, a population for which treatment personalization is particularly challenging.
The results were described as demonstrating 'the feasibility of generating personalized atrial digital twins that reproduce patient-specific activation patterns and partially capture activation-based EGM biomarker trends.'
What This Means
This research suggests that it is possible to create personalized computer models of individual patients' hearts — called 'digital twins' — that accurately mimic how electrical signals travel through the atria (the upper chambers of the heart) in people with persistent atrial fibrillation (AF), a common and difficult-to-treat heart rhythm disorder. The researchers built these models for 20 patients using heart scans (CT images) and electrical recordings taken during clinical procedures, then fine-tuned the models step by step until the simulated electrical patterns closely matched what was actually measured in patients. After this calibration process, the average error in predicting when different parts of the heart activate dropped from about 36 milliseconds to about 14 milliseconds, a significant improvement.
The digital twins were also able to reproduce broad patterns seen in clinical electrical recordings (called electrograms), such as differences in signal complexity and timing variability between healthy and diseased heart tissue. However, the models were less accurate when it came to reproducing the actual size (voltage amplitude) of the electrical signals, meaning they captured the 'shape' of abnormalities but not their full magnitude. This points to an area where the models still need improvement before they could be considered fully realistic representations of clinical recordings.
This research matters because persistent AF is notoriously hard to treat — current medications and ablation procedures (which destroy small areas of heart tissue to stop abnormal electrical circuits) do not work well for many patients. By developing accurate personalized heart models, doctors and researchers could one day simulate different treatment strategies on a patient's digital twin before performing the actual procedure, helping to identify the most effective approach for each individual. While the current models are not yet ready for direct clinical use due to limitations in voltage reproduction, this study represents a meaningful step toward that goal.
de Luis-Moura D, Celotto C, Golmaryami S, Termenón-Rivas M, Romitti G, Moreno-Pineda S, et al.. (2026). Atrial digital twins reproducing clinical biomarkers of intracardiac electrograms.. Computers in biology and medicine. https://doi.org/10.1016/j.compbiomed.2026.111908