Performance and limitations of noninvasive cardiac activation mapping
Abstract
BACKGROUND: Activation mapping using noninvasive electrocardiographic imaging (ECGi) has recently been used to describe the physiology of different cardiac abnormalities. These descriptions differ from prior invasive studies, and both methods have not been thoroughly confronted in a clinical setting. OBJECTIVE: The goal of the present study was to provide validation of noninvasive activation mapping in a clinical setting through direct confrontation with invasive epicardial contact measures. METHODS: Fifty-nine maps were obtained in 55 patients and aligned on a common geometry. Nearest-neighbor interpolation was used to avoid map smoothing. Quantitative comparison was performed by computing between-map correlation coefficients and absolute activation time errors. RESULTS: The mean activation time error was 20.4 ± 8.6 ms, and the between-map correlation was poor (0.03 ± 0.43). The results suggested high interpatient variability (correlation -0.68 to 0.82), wide QRS patterns, and paced rhythms demonstrating significantly better mean correlation (0.68 ± 0.17). Errors were greater in scarred regions (21.9 ± 10.8 ms vs 17.5 ± 6.7 ms; P < .01). Fewer epicardial breakthroughs were imaged using noninvasive mapping (1.3 ± 0.5 vs 2.3 ± 0.7; P < .01). Primary breakthrough locations were imaged 75.7 ± 38.1 mm apart. Lines of conduction block (jumps of ≥50 ms between contiguous points) due to structural anomalies were recorded in 27 of 59 contact maps and were not visualized at these same sites noninvasively. Instead, artificial lines appeared in 33 of 59 noninvasive maps in regions of reduced bipolar voltage amplitudes (P = .03). An in silico model confirms these artificial constructs. CONCLUSION: Overall, agreement of ECGi activation mapping and contact mapping is poor and heterogeneous. The between-map correlation is good for wide QRS patterns. Lines of block and epicardial breakthrough sites imaged using ECGi are inaccurate. Further work is required to improve the accuracy of the technique.
Related
- Multi-modal AI: Integrating electrocardiography with molecular biomarkers for noninvasive detection of cardiac allograft rejection
- Automated and noninvasive in vivo mapping of brain functions in the human cortex
- Noninvasive brain stimulation for migraine : A bibliometric analysis of scientific research with thematic and keyword mapping
- Clinical performance and practical limitations of large language models in respiratory medicine
- Can LLMs Succeed in National Dental Examinations? A Comprehensive Analysis of Performance, Strengths, and Limitations
BibTeX
@article{duchateau2026performance,
title = {{Performance and limitations of noninvasive cardiac activation mapping}},
author = {Josselin Duchateau and Frédéric Sacher and Thomas Pambrun and Nicolas Derval and Judit Chamorro-Servent and Arnaud Denis and Sylvain Ploux and Mélèze Hocini and Pierre Jaïs and Olivier Bernus and Michel Haïssaguerre and Rémi Dubois},
journal = {Heart Rhythm},
year = {2026},
month = oct,
doi = {10.1016/j.hrthm.2018.10.010},
url = {https://hal.science/hal-05796015v1},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Performance and limitations of noninvasive cardiac activation mapping
AU - Josselin Duchateau
AU - Frédéric Sacher
AU - Thomas Pambrun
AU - Nicolas Derval
AU - Judit Chamorro-Servent
AU - Arnaud Denis
AU - Sylvain Ploux
AU - Mélèze Hocini
AU - Pierre Jaïs
AU - Olivier Bernus
AU - Michel Haïssaguerre
AU - Rémi Dubois
PY - 2026
DA - 2026/10/12
JO - Heart Rhythm
DO - 10.1016/j.hrthm.2018.10.010
UR - https://hal.science/hal-05796015v1
AB - BACKGROUND: Activation mapping using noninvasive electrocardiographic imaging (ECGi) has recently been used to describe the physiology of different cardiac abnormalities. These descriptions differ from prior invasive studies, and both methods have not been thoroughly confronted in a clinical setting. OBJECTIVE: The goal of the present study was to provide validation of noninvasive activation mapping in a clinical setting through direct confrontation with invasive epicardial contact measures. METHODS: Fifty-nine maps were obtained in 55 patients and aligned on a common geometry. Nearest-neighbor interpolation was used to avoid map smoothing. Quantitative comparison was performed by computing between-map correlation coefficients and absolute activation time errors. RESULTS: The mean activation time error was 20.4 ± 8.6 ms, and the between-map correlation was poor (0.03 ± 0.43). The results suggested high interpatient variability (correlation -0.68 to 0.82), wide QRS patterns, and paced rhythms demonstrating significantly better mean correlation (0.68 ± 0.17). Errors were greater in scarred regions (21.9 ± 10.8 ms vs 17.5 ± 6.7 ms; P < .01). Fewer epicardial breakthroughs were imaged using noninvasive mapping (1.3 ± 0.5 vs 2.3 ± 0.7; P < .01). Primary breakthrough locations were imaged 75.7 ± 38.1 mm apart. Lines of conduction block (jumps of ≥50 ms between contiguous points) due to structural anomalies were recorded in 27 of 59 contact maps and were not visualized at these same sites noninvasively. Instead, artificial lines appeared in 33 of 59 noninvasive maps in regions of reduced bipolar voltage amplitudes (P = .03). An in silico model confirms these artificial constructs. CONCLUSION: Overall, agreement of ECGi activation mapping and contact mapping is poor and heterogeneous. The between-map correlation is good for wide QRS patterns. Lines of block and epicardial breakthrough sites imaged using ECGi are inaccurate. Further work is required to improve the accuracy of the technique.
SN - 1547-5271
SN - 1556-3871
ER -