Epicardial Wave Mapping in Persistent Atrial Fibrillation
Epicardial Wave Mapping in Persistent Atrial Fibrillation
This study included 20 patients with PerAF (defined in accordance with HRS Expert Consensus Statement) undergoing elective cardiac surgery for mitral regurgitation (12), aortic stenosis (2), aortic regurgitation (1), hypertrophic obstructive cardiomyopathy (1), mitral stenosis (1), and ischaemic heart disease (3). All the patients gave written and informed consent prior to the surgery and the study protocol was approved by the research and ethics committee of Melbourne Health.
Following median sternotomy and prior to cardiopulmonary bypass, high-density atrial epicardial mapping was performed. A custom-made high-density triangular epicardial plaque comprising 128 silver-plated copper electrodes with an inter-electrode distance of 2.5 mm (effective mapping area 6.75 cm) was positioned sequentially at four regions (i) oblique sinus on the posterior left atrium (PLAW) between the pulmonary veins, (ii) most anterior surface (surface which is in direct contact with the visceral pericardium) of the left atrial appendage (LAA), (iii) most anterior surface (surface which is in direct contact with the visceral pericardium) of the right atrial appendage (RAA), (iv) the anterior surface of the right superior RSPV-LA junction (RSPV-LA). Electrograms were sampled at 1000 Hz with a band pass filter of 0.05 to 400 Hz employed. Continuous bipolar atrial electrograms were recorded using a computerized mapping system (Unemap, Uniservices, Auckland, New Zealand) for offline analysis. Ten second recordings were obtained after plaque stability and signal quality had been confirmed.
Beat-to-Beat Activation Time The entire recording was scanned to ensure overall signal quality. Unemap signals were imported into the customized computer software Cardiac ElectroPhysiology Analysis System (CEPAS, Cuoretech Pty Ltd, Sydney, Australia) that was used to determine beat-to-beat activation times from each bipole on the epicardial mapping plaque. CEPAS has specific user-defined characteristics to identify electrogram activations. These include (i) a baseline noise threshold; (ii) electrogram width criterion to avoid detection of broad far-field activations; (iii) electrogram slope; and (iv) electrogram 'refractory' periods to avoid multiple detections within the same activation. Based on the previous work by Aizer et al., a noise threshold of 0.1 mV, width criterion of 10 ms, and refractory period of 50 ms were utilized for the analysis. All automated electrogram analysis was visually verified to ensure accurate annotation of activation times. Activations were manually corrected if automated annotation was incorrect (Figure 1).
(Enlarge Image)
Figure 1.
Annotation of electrograms. (A) An example of the automated electrogram annotation of non-fractionated electrograms using Cardiac ElectroPhysiology Analysis System software. The algorithm was set to mark the local maximum deflection. A red dot shows the location of the automated annotation used to calculate the beat-to-beat AFCL. All automated annotation was manually verified and when automated annotation was incorrect (yellow arrows) manually correction was undertaken. (B) An example of a site with discrete multi-component electrograms (>3 deflections of >50 ms duration) separated by a discrete isoelectric baseline. These sites were manually annotated to the onset of the local electrogram activation. (C) A site with continuous electrical activity-complex-fractionated atrial electrogram. Given the ambiguity of annotating activation times, these sites were excluded from the animation and the AFCL analysis. The location of these sites was recorded for spatial analysis.
The first step in our analysis was to determine the beat-to-beat activation time from each bipolar recording site. Thus for the purposes of this analysis each bipole site was defined by two specific variables (i) the Cartesian co-ordinates for the bipole; (ii) a time series of activation times (t) referenced to the onset of the 10 s recording (time zero) for each atrial depolarization. This data set was then exported in the matrix form for wavefront animation and analysis.
Complex-fractionated Atrial Electrograms Complex-fractionated atrial electrograms (CFAEs) were defined as electrograms displaying continuous electrical activity (CEA) over the entire 10 s recording period consistent with the original description. The location of CFAE sites was recorded for spatial analysis. Given the ambiguity of assigning activation times at sites of CEA, these sites were excluded from the animation and atrial fibrillation cycle length (AFCL) analysis.
Multi-component electrograms were defined by ≥3 deflections over >50 ms duration separated by a discrete iso-electric baseline. These were manually annotated at the onset of electrogram activation. Figure 1 shows representative examples of annotation of non-fractionated, multi-component, and CFAE electrograms from the data set.
AF Mapping Within the 10 s recording, each bipolar site was sampled at 1 ms intervals (sampling rate of mapping system of 1000 Hz). Using commercially available software (DataTank, Visual Data Tools, Inc., Chapel Hill, NC, USA), individual activations at each fixed bipolar site on the epicardial plaque were animated independently.
Definition of Timing Intervals
Activation Time Window When activation occurred at a given site, this was animated 'on' for a duration of 20 ms being the activation time window. Atrial slow conduction has been defined as a local conduction velocity of 10–20 cm/s and conduction block as <10 cm/s. Given the 2.5 mm inter-bipole spacing between adjacent bipoles, a difference in local activation times of 25 and 35 ms (horizontal and oblique trajectories) between adjacent bipoles would represent conduction block. It is unlikely that two adjacent bipoles are activated by the leading edge of the same wavefront if the difference in local activation times were >25 ms in the horizontal direction or 35 ms in the oblique direction. These values set the physiological limits of normal wavefront propagation and were used for the animation process. Using heuristic principles, Activation Time Windows from 1 to 35 ms were analysed and an activation time window of 20 ms resulted in the best discrimination of wavefront patterns. However, the evaluation of activation time windows over a range of 15–35 ms did not result in differences in map interpretation. The activation time window determined whether activations at adjacent bipoles were displayed either (i) simultaneously and as part of the leading edge of the same depolarizing wavefront; (ii) sequentially as propagation of the leading edge of an existing wavefront or (iii) discontinuously and activated as part of a new wavefront. In a recent epicardial mapping study, de Groot et al. used similar values (15–40 ms) when defining discontinuous conduction as fibrillation waves starting from the boundary of another wave. All maps were analysed at a speed of 20 ms/s. When interpretation needed further clarification, the play back speed was reduced to 5 ms/s.
Refractory Period The refractory period of a given bipole was defined at 50 ms. Repeated activations occurring within this interval were either part of a multi-component signal or CFAE or represented far-field activity.
Dominant frequency (DF) analysis was performed using CEPAS on the raw electrogram signals exported from the Unemap data. Within the same 10-s recording window, spectral analysis for determination of DF was analysed. For the specific purposes of DF analysis, exported signals were rectified, filtered using a Butterworth filter and edge-tapered with a Hanning window. Dominant frequency was determined by Fast Fourier Transform using zero-padding with a spectral resolution of 0.1 Hz. The DF of each individual bipole-recording site was defined as the frequency demonstrating the highest power within the 3–15-Hz frequency domain. For the purposes of the spatial analysis in this study, a high-dominant frequency (HDF) site within the recording plaque was identified if the local bipole DF was 20% greater than the DF of its adjacent bipoles.
Based on prior studies, activation patterns were classified into four morphologies; wavefronts, focal activations, rotational circuits, and disorganized activity (Supplementary material online, Movies S1–4).
The number, size, and direction of wavefront propagation were noted for each activation. We measured the size of wavefronts that were visible within the mapping field. Wavefronts that only appear at the edges of the mapping field or activate the entire mapping area could not be determined. Wavefront directionality was determined using visual inspection. Wavefronts were subclassified according to the following definitions:
Focal Activation Earliest activation occurs at a discrete bipole site located inside the mapping area with centrifugal wavefront activation that spreads radially to the periphery.
Rotational Circuit A rotational wave (≥2 rotations of 360°) of excitation centred on a central area located inside the mapping area.
Disorganized Activity Activations that do not fulfil the criteria for a wavefront and are composed of early activation at >2 adjacent bipoles inside the mapping area that propagate ≤3 bipoles or activations that occur as isolated beats dissociated from the activation of adjacent bipole sites. 'Disorganized' activations are characterized by multiple simultaneous 'epicardial breakthroughs' within the mapping field with collision of wavefronts so that there is no obvious organized pattern of activation.
Preferential Conduction Pathway Preferential conduction pathway was defined by the presence of a stereotypical repeated pattern (>2) and trajectory of wavefront activation within the mapping area during the recording period. Preferential conduction was determined by the visual analysis of the AF maps.
All statistical analysis was performed using SPSS software version 17.0 (SPSS, Chicago, IL, USA) Normality of all quantitative data variables was checked using the Kolmogorov–Smirnov test. Continuous variables are reported as means ± standard deviation and median and inter-quartile range (IQR), as appropriate. Categorical variables are reported as numbers and percentages. Comparisons of continuous variables between different anatomic regions were performed using a one-way analysis of variance, with post hoc analysis using Bonferroni correction for multiple comparisons.
To determine the degree of intra- and inter-observer variability in activation pattern classification, the principal investigator and a second observer (S.K.) were asked to classify a blinded sample of 100 patterns from the data set using the above study criteria. For the principal investigator, there was a 98% INTRA-observer agreement in the classification of activation patterns, kappa = 0.95, P < 0.001. For the second observer (S.K.), there was 97% INTER-observer agreement with the principal investigator in the classification of activation patterns, kappa = 0.86, P < 0.05. In cases where there was disagreement between the principal and the second observer, activation patterns were decided after review and mutual consensus. All tests were two-sided and a P-value <0.05 was considered statistically significant.
Methods
This study included 20 patients with PerAF (defined in accordance with HRS Expert Consensus Statement) undergoing elective cardiac surgery for mitral regurgitation (12), aortic stenosis (2), aortic regurgitation (1), hypertrophic obstructive cardiomyopathy (1), mitral stenosis (1), and ischaemic heart disease (3). All the patients gave written and informed consent prior to the surgery and the study protocol was approved by the research and ethics committee of Melbourne Health.
Data Acquisition
Following median sternotomy and prior to cardiopulmonary bypass, high-density atrial epicardial mapping was performed. A custom-made high-density triangular epicardial plaque comprising 128 silver-plated copper electrodes with an inter-electrode distance of 2.5 mm (effective mapping area 6.75 cm) was positioned sequentially at four regions (i) oblique sinus on the posterior left atrium (PLAW) between the pulmonary veins, (ii) most anterior surface (surface which is in direct contact with the visceral pericardium) of the left atrial appendage (LAA), (iii) most anterior surface (surface which is in direct contact with the visceral pericardium) of the right atrial appendage (RAA), (iv) the anterior surface of the right superior RSPV-LA junction (RSPV-LA). Electrograms were sampled at 1000 Hz with a band pass filter of 0.05 to 400 Hz employed. Continuous bipolar atrial electrograms were recorded using a computerized mapping system (Unemap, Uniservices, Auckland, New Zealand) for offline analysis. Ten second recordings were obtained after plaque stability and signal quality had been confirmed.
Data Analysis
Beat-to-Beat Activation Time The entire recording was scanned to ensure overall signal quality. Unemap signals were imported into the customized computer software Cardiac ElectroPhysiology Analysis System (CEPAS, Cuoretech Pty Ltd, Sydney, Australia) that was used to determine beat-to-beat activation times from each bipole on the epicardial mapping plaque. CEPAS has specific user-defined characteristics to identify electrogram activations. These include (i) a baseline noise threshold; (ii) electrogram width criterion to avoid detection of broad far-field activations; (iii) electrogram slope; and (iv) electrogram 'refractory' periods to avoid multiple detections within the same activation. Based on the previous work by Aizer et al., a noise threshold of 0.1 mV, width criterion of 10 ms, and refractory period of 50 ms were utilized for the analysis. All automated electrogram analysis was visually verified to ensure accurate annotation of activation times. Activations were manually corrected if automated annotation was incorrect (Figure 1).
(Enlarge Image)
Figure 1.
Annotation of electrograms. (A) An example of the automated electrogram annotation of non-fractionated electrograms using Cardiac ElectroPhysiology Analysis System software. The algorithm was set to mark the local maximum deflection. A red dot shows the location of the automated annotation used to calculate the beat-to-beat AFCL. All automated annotation was manually verified and when automated annotation was incorrect (yellow arrows) manually correction was undertaken. (B) An example of a site with discrete multi-component electrograms (>3 deflections of >50 ms duration) separated by a discrete isoelectric baseline. These sites were manually annotated to the onset of the local electrogram activation. (C) A site with continuous electrical activity-complex-fractionated atrial electrogram. Given the ambiguity of annotating activation times, these sites were excluded from the animation and the AFCL analysis. The location of these sites was recorded for spatial analysis.
The first step in our analysis was to determine the beat-to-beat activation time from each bipolar recording site. Thus for the purposes of this analysis each bipole site was defined by two specific variables (i) the Cartesian co-ordinates for the bipole; (ii) a time series of activation times (t) referenced to the onset of the 10 s recording (time zero) for each atrial depolarization. This data set was then exported in the matrix form for wavefront animation and analysis.
Complex-fractionated Atrial Electrograms Complex-fractionated atrial electrograms (CFAEs) were defined as electrograms displaying continuous electrical activity (CEA) over the entire 10 s recording period consistent with the original description. The location of CFAE sites was recorded for spatial analysis. Given the ambiguity of assigning activation times at sites of CEA, these sites were excluded from the animation and atrial fibrillation cycle length (AFCL) analysis.
Multi-component electrograms were defined by ≥3 deflections over >50 ms duration separated by a discrete iso-electric baseline. These were manually annotated at the onset of electrogram activation. Figure 1 shows representative examples of annotation of non-fractionated, multi-component, and CFAE electrograms from the data set.
AF Mapping Within the 10 s recording, each bipolar site was sampled at 1 ms intervals (sampling rate of mapping system of 1000 Hz). Using commercially available software (DataTank, Visual Data Tools, Inc., Chapel Hill, NC, USA), individual activations at each fixed bipolar site on the epicardial plaque were animated independently.
Definition of Timing Intervals
Activation Time Window When activation occurred at a given site, this was animated 'on' for a duration of 20 ms being the activation time window. Atrial slow conduction has been defined as a local conduction velocity of 10–20 cm/s and conduction block as <10 cm/s. Given the 2.5 mm inter-bipole spacing between adjacent bipoles, a difference in local activation times of 25 and 35 ms (horizontal and oblique trajectories) between adjacent bipoles would represent conduction block. It is unlikely that two adjacent bipoles are activated by the leading edge of the same wavefront if the difference in local activation times were >25 ms in the horizontal direction or 35 ms in the oblique direction. These values set the physiological limits of normal wavefront propagation and were used for the animation process. Using heuristic principles, Activation Time Windows from 1 to 35 ms were analysed and an activation time window of 20 ms resulted in the best discrimination of wavefront patterns. However, the evaluation of activation time windows over a range of 15–35 ms did not result in differences in map interpretation. The activation time window determined whether activations at adjacent bipoles were displayed either (i) simultaneously and as part of the leading edge of the same depolarizing wavefront; (ii) sequentially as propagation of the leading edge of an existing wavefront or (iii) discontinuously and activated as part of a new wavefront. In a recent epicardial mapping study, de Groot et al. used similar values (15–40 ms) when defining discontinuous conduction as fibrillation waves starting from the boundary of another wave. All maps were analysed at a speed of 20 ms/s. When interpretation needed further clarification, the play back speed was reduced to 5 ms/s.
Refractory Period The refractory period of a given bipole was defined at 50 ms. Repeated activations occurring within this interval were either part of a multi-component signal or CFAE or represented far-field activity.
Dominant Frequency
Dominant frequency (DF) analysis was performed using CEPAS on the raw electrogram signals exported from the Unemap data. Within the same 10-s recording window, spectral analysis for determination of DF was analysed. For the specific purposes of DF analysis, exported signals were rectified, filtered using a Butterworth filter and edge-tapered with a Hanning window. Dominant frequency was determined by Fast Fourier Transform using zero-padding with a spectral resolution of 0.1 Hz. The DF of each individual bipole-recording site was defined as the frequency demonstrating the highest power within the 3–15-Hz frequency domain. For the purposes of the spatial analysis in this study, a high-dominant frequency (HDF) site within the recording plaque was identified if the local bipole DF was 20% greater than the DF of its adjacent bipoles.
Classification of Activation Pattern Morphologies
Based on prior studies, activation patterns were classified into four morphologies; wavefronts, focal activations, rotational circuits, and disorganized activity (Supplementary material online, Movies S1–4).
Wavefronts
The number, size, and direction of wavefront propagation were noted for each activation. We measured the size of wavefronts that were visible within the mapping field. Wavefronts that only appear at the edges of the mapping field or activate the entire mapping area could not be determined. Wavefront directionality was determined using visual inspection. Wavefronts were subclassified according to the following definitions:
Broad wavefronts: where the width of the activation wave front spans the entire plaque, activating it in a linear fashion.
One narrow WF: the width of the propagating wavefront is only 2–6 bipoles wide.
Two narrow WFs: two narrow wavefronts occupy the mapping area at the same time.
Three narrow WFs: ≥3 narrow wavefronts occupy the mapping area at the same time.
Focal Activation Earliest activation occurs at a discrete bipole site located inside the mapping area with centrifugal wavefront activation that spreads radially to the periphery.
Rotational Circuit A rotational wave (≥2 rotations of 360°) of excitation centred on a central area located inside the mapping area.
Disorganized Activity Activations that do not fulfil the criteria for a wavefront and are composed of early activation at >2 adjacent bipoles inside the mapping area that propagate ≤3 bipoles or activations that occur as isolated beats dissociated from the activation of adjacent bipole sites. 'Disorganized' activations are characterized by multiple simultaneous 'epicardial breakthroughs' within the mapping field with collision of wavefronts so that there is no obvious organized pattern of activation.
Preferential Conduction Pathway Preferential conduction pathway was defined by the presence of a stereotypical repeated pattern (>2) and trajectory of wavefront activation within the mapping area during the recording period. Preferential conduction was determined by the visual analysis of the AF maps.
Statistical Analysis
All statistical analysis was performed using SPSS software version 17.0 (SPSS, Chicago, IL, USA) Normality of all quantitative data variables was checked using the Kolmogorov–Smirnov test. Continuous variables are reported as means ± standard deviation and median and inter-quartile range (IQR), as appropriate. Categorical variables are reported as numbers and percentages. Comparisons of continuous variables between different anatomic regions were performed using a one-way analysis of variance, with post hoc analysis using Bonferroni correction for multiple comparisons.
To determine the degree of intra- and inter-observer variability in activation pattern classification, the principal investigator and a second observer (S.K.) were asked to classify a blinded sample of 100 patterns from the data set using the above study criteria. For the principal investigator, there was a 98% INTRA-observer agreement in the classification of activation patterns, kappa = 0.95, P < 0.001. For the second observer (S.K.), there was 97% INTER-observer agreement with the principal investigator in the classification of activation patterns, kappa = 0.86, P < 0.05. In cases where there was disagreement between the principal and the second observer, activation patterns were decided after review and mutual consensus. All tests were two-sided and a P-value <0.05 was considered statistically significant.
Source...