Related Experiment Video
Updated: Jul 19, 2026
High-Resolution Endocardial and Epicardial Optical Mapping in a Sheep Model of Stretch-Induced Atrial Fibrillation
Published on: July 29, 2011
Automatic mapping of human atrial fibrillation by template matching
Richard P M Houben1, Natasja M S de Groot, Fred W Lindemans
1Medtronic Bakken Research Center, Maastricht, The Netherlands.
Background:
The high spatiotemporal variation in morphology of fibrillation electrograms makes mapping of atrial fibrillation (AF) a difficult and burdensome task.
Objectives:
The purpose of this study was to evaluate the results of automatic detection of fibrillation electrograms by a template matching technique.
Methods:
During cardiac surgery in 25 patients without a history of AF, paroxysms of AF were induced by rapid atrial pacing. A mapping array of 244 unipolar electrodes (3.6-cm diameter, 2.25-mm interelectrode distance) was positioned on the free wall of the right atrium. All fibrillation electrograms were correlated with a mathematically constructed library of 128 potentials of different duration, RS ratio, and short double components. The moments of maximal correlation, coinciding with the negative deflection in the fibrillation potentials, were used to create fibrillation maps.
Results:
In each patient, a segment of 18.6 +/- 3.8 seconds of AF was analyzed, resulting in 80 to 130 maps per patient. The output of the automatic algorithm was compared with careful manual analysis by an experienced investigator. Of the total database of 398,796 fibrillation potentials, 93.6% +/- 4.2% resulted in a good correlation with one of the templates in the library (correlation coefficient >= 0.7). At a correlation threshold of 0.6, on average template matching yielded slightly more false-positive than false-negative detections (sensitivity 96.6% +/- 2.5%, positive predictive value 94.3% +/- 5.4%). The majority of false-positive detections were due to electrotonic potentials recorded along the lateral boundaries of the fibrillation waves. This led to a slight overlap of fibrillation waves but not to false detection of nonexisting wavefronts. Undersensing was mainly due to the presence of long double and fractionated potentials (2.6%) that were not represented in the template library. Fractionated parts in the electrograms were identified by failure of template matching and can be analyzed separately.
Conclusion:
Template matching is a useful technique for characterizing unipolar fibrillation electrograms and for visualizing the complex activation patterns during AF. It allows automatic evaluation of the electropathologic substrate of AF on an individual basis.
