The dominant morphology of fractionated atrial electrograms has greater temporal stability in persistent as compared with paroxysmal atrial fibrillation

Computers in Biology and Medicine 43 (2013) 2127–2135

Contents lists available at ScienceDirect

Computers in Biology and Medicine journal homepage: www.elsevier.com/locate/cbm

The dominant morphology of fractionated atrial electrograms has greater temporal stability in persistent as compared with paroxysmal atrial fibrillation Edward J. Ciaccio n, Angelo B. Biviano, Hasan Garan Department of Medicine, Division of Cardiology, Columbia University Medical Center, NY, USA

art ic l e i nf o

a b s t r a c t

Article history: Received 26 October 2012 Accepted 16 August 2013

Background: Measurements of both the dominant frequency (DF) and the time series morphology of complex fractionated atrial electrograms (CFAE) are useful to distinguish persistent from paroxysmal atrial fibrillation (AF). In this study, an algorithm was devised to extract morphologic components according to frequency, and its usefulness for distinguishing CFAE was shown. Method: CFAE of length 16 s were obtained at two sites each from the four pulmonary vein ostia (PV), and from anterior and posterior left atrial free wall (FW), in nine paroxysmal and 10 longstanding persistent AF patients. The DF was computed for each of two 8 s CFAE segments in each 16 s recording. Each CFAE segment was then transformed into a set of basis vectors, which represent electrogram morphology at each frequency. The dominant morphology (DM) is defined as the ensemble average of sequential signal segments, with the segment length equal to the period at the DF. The DMs of the two 8 s pairs were correlated. Normalized correlation coefficients were tabulated for all data, and separately for PV and FW. The means and coefficients of variation of the DM correlation coefficients were then plotted, and a linear discriminant function was used to classify persistent versus paroxysmal AF data. For comparison with DM results, CFE-mean and interval confidence level (ICL) were also calculated for persistent versus paroxysmal AF data. Results: Mean correlation of the DM, 1st 8 s versus 2nd 8 s data, was 0.62 þ0.22 for persistent versus 0.50 þ0.19 for paroxysmal CFAE for all recording sites (po0.001). At single anatomical locations, correlation was greater in persistents than paroxysmals at all sites, but achieved significance only at the left superior (po 0.001) and right superior (po 0.05) PV. Spatial variation in correlation coefficient was greater in paroxysmal than persistent AF (not significant). Using the means of DF correlation coefficients, 17/19 patients were classified correctly. The CFE-mean parameter averaged 89.01 720.99 ms in persistents versus 93.96 7 33.81 ms in paroxysmals (p o0.05), while ICL averaged 94.54 7 18.52 deflections/8 s for persistents versus 90.707 19.28 deflections/8 s for paroxysmals (po 0.05). Conclusions: In CFAE recordings, the DM parameter was found to have greater temporal morphologic variation in paroxysmal as compared with persistent AF data (po 0.001). In contrast, only moderate significance between paroxysmal versus persistent AF data was found when using the of CFE-mean and ICL parameters (po 0.05). The DM parameter may thus be useful as a new measure to discern both temporal and spatial variations in CFAE in paroxysmal versus persistent AF recordings. & 2013 Elsevier Ltd. All rights reserved.

Keywords: Atrial fibrillation Dominant frequency Dominant morphology Fractionation Spectral analysis

1. Background Isolation of electrical activity in the pulmonary veins (PV) is a first step to prevent atrial fibrillation (AF) when drug therapy fails [1,2]. This technique as a sole procedure works reasonably well in patients with paroxysmal AF [3]. Ablation of other areas of the left n Correspondence to: Harkness Pavilion 934, 180 Fort Washington Avenue, Columbia University, New York, NY 10032, USA. Tel.: þ1 212 305 5447; fax: þ1 212 342 0447. E-mail address: [email protected] (E.J. Ciaccio).

0010-4825/$ - see front matter & 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.compbiomed.2013.08.027

atrium may eliminate arrhythmogenic sites, stop AF, and prevent its recurrence in both paroxysmal [4] and persistent AF substrates [5], although additional procedures may be required, particularly in cases of persistent AF [6]. Sites with complex fractionated atrial electrograms (CFAE) have been proposed as arrhythmogenic targets [7] for catheter ablation. CFAE are defined as electrograms with continuous electrical activity without isoelectric segment 450 ms, or activity with period o100 ms [7]. Since areas of the left atrium containing CFAE may be extensive, ablating all of the CFAE sites in AF patients can markedly increase procedure time and can possibly cause morbidity. Thus it may be useful to

2128

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

characterize CFAE by quantitative means for detection of specific characteristics that can be helpful toward deciding target sites for ablation. Currently, the dominant frequency (DF) is a ubiquitous tool to quantitatively characterize CFAE during and after the electrophysiologic study of AF patients [8,9]. It can be defined as the largest fundamental component of the frequency spectrum within the electrophysiologic range of interest [8]. The DF was computed in an early work by preprocessing the signal with a bandpass filter, followed by rectification and low pass filtering [10,11]. In a more recent work, the preprocessing stage was eliminated to prevent signal distortion [12]. Spectral estimators which are more robust to random and phase noise as compared with the Fourier analysis have also recently been devised [13]. Furthermore, spectral parameters besides the DF have been developed as additional measures of the frequency characteristics of CFAE [13]. These include the dominant amplitude (DA), which is defined as the amplitude of the DF spectral peak. The parameter is related to the power under the dominant peak, but unlike other methods, does not require guestimation of the start and end of the peak [13]. A smaller value of DA indicates less power in the dominant peak, more power in the background level, and therefore greater complexity of electrical activity. The mean and standard deviation in the spectral profile (MP and SP) have been recently described [13]. These parameters also do not require guestimation concerning the dominant peak or its harmonics. A greater value of MP and/or SP indicates a higher background level and therefore greater complexity of electrical activity. Prior findings of lesser DA and greater MP and SP in paroxysmal AF [13] suggest that there is a greater level of instability in the atrial activation pattern during paroxysmal as compared with persistent AF. This finding represented a first step in quantitatively characterizing differences in the substrate between these AF types. A recent work has also found morphologic differences in the time series of these signals to be useful in characterizing paroxysmal versus persistent AF [14,15]. When the morphologic descriptors are more variable, it is indicative of increasing instability in the electrical activation pattern from which the extracellular signals are formed. More variable morphologic descriptors, and therefore increased instability, were again found in paroxysmal as compared with persistent AF recordings [14,15]. Therefore, both frequency and morphologic measurements are suggestive that significant differences in the electrical activation pattern exist in paroxysmal versus persistent AF, and that the persistent AF recordings are less variable and more stable, possibly due to the presence of relatively intransigent drivers. If frequency and morphologic modalities could be combined for analysis of CFAE, it would potentially provide a robust means for analysis by taking into account more information about the electrogram signals being analyzed. Herein, we describe a method to characterize CFAE using both morphologic descriptors and frequency analysis. Since AF drivers of electrical activity may be present at several frequencies, the new method is potentially useful to identify and characterize multiple distinct drivers of this arrhythmia.

2. Method 2.1. Clinical data acquisition and electrophysiologic mapping Atrial electrograms were recorded from 19 patients referred to the Columbia University Medical Center cardiac electrophysiology laboratory for catheter ablation of AF. These recordings were obtained prospectively as approved by the Institutional Review Board, and analyzed retrospectively for this study. Nine patients had documented clinical paroxysmal AF. Normal sinus rhythm was

their baseline cardiac rhythm in the electrophysiology laboratory. Induction of AF was done via burst atrial pacing from the coronary sinus or right atrial lateral wall, and allowed to persist for at least 10 min prior to signal acquisition. Ten other patients had longstanding persistent AF without interruption for several months to many years prior to catheter mapping and ablation. The bipolar atrial mapping procedure was performed with a NaviStar ThermoCool catheter, 7.5 F, 3.5 mm tip, with 2 mm spacing between bipoles (Biosense-Webster Inc., Diamond Bar, CA, USA). The electrogram signals were acquired using a General Electric CardioLab system (GE Healthcare, Waukesha, WI), and filtered at acquisition from 30 to 500 Hz with a single-pole bandpass filter to remove baseline drift and high frequency noise. The filtered signals were digitally sampled at 977 Hz and the digital data was then stored. Although the high end of the bandpass filter was slightly above the Nyquist frequency, negligible CFAE signal energy resides in this frequency range [16]. CFAE were identified by two clinical cardiac electrophysiologists using the definition provided in Section 1 and described in detail elsewhere [7]. CFAE recordings of at least 16 s in duration were obtained from two sites each outside the ostia of the four pulmonary veins. Similar recordings were obtained at two left atrial free wall sites, one in the mid posterior wall, and another on the anterior ridge at the base of the left atrial appendage. The mapping catheter was navigated in these pre-specified areas until a CFAE site was identified. A total of 204 CFAE sequences – 90 from paroxysmal and 114 from longstanding AF patients were included in the following quantitative analysis. As in previous studies, to standardize the morphological characteristics, all CFAE were normalized to mean zero and unity variance (average level ¼0 V, standard deviation and variance ¼1) [13,17]. 2.2. Construction and correlation of the DM Ensemble average-type power spectra were generated, as described in detail elsewhere [13,17]. The ensemble-type spectra have doubled frequency resolution in the physiologic range of 3–12 Hz as compared to the Fourier analysis [18]. A linear transformation derived from ensemble-type spectral analysis was used to decompose the CFAE segments and form data-driven, orthogonal basis vectors [17]. The transform equation is a N ðwÞ ¼ n ¼ int

1 T NN ðwÞ  x N n N w

ð1Þ ð2Þ

where aw are the basis vectors, N is a dimension, ‘int’ is the integer function, and the transformation matrix is given by 2 3 Iw Iw ⋯ Iw 6 Iw Iw ⋯ Iw 7 6 7 T NN ðwÞ ¼ 6 ð3Þ 7 4⋯ ⋯ ⋯ ⋯5 Iw Iw ⋯ Iw where Iw is the identity matrix of dimension w  w. The transformation matrix Tw is used to construct repeating ensemble averages ew from successive segments of signal xN, with each ensemble average having a length w. The basis is orthogonal for values of segment length w lacking integer relationships. For this study, basis vectors were constructed for w¼wn, where wn ¼

sample rate DF

ð4Þ

with the sample rate of the acquisition system being 977 Hz. The dominant morphology (DM) is defined as the ensemble average ewn of sequential signal segments, with the segment length equal

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

to the period at the DF [17]. The basis vector a N ðwn Þ is ewn repeated to a length N. The DM of the two 8 s pairs for each 16 s CFAE recording was correlated. The 1st and 2nd 8 s DMs will differ when the DFs of each differ, and even when the DF values are the same, there will be differences because of changes in the signal used to form each ensemble average. The normalized correlation coefficient is given by e 1 ð1 : wn Þ  e 2 ð1 : wn Þ ffi cc ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi n ½ðe 1 ð1 : w Þ  e 1 ð1 : wn ÞÞðe 2 ð1 : wn Þ  e 2 ð1 : wn Þ

ð5Þ

where e1 is the ensemble average of the 1st 8 s segment at its DF, e2 is the ensemble average of the 2nd 8 s segment at its DF, wn is the ensemble average length for the 1st 8 s segment, and ‘∙’ denotes the inner product. To compute cc, e2 is concatenated when it is longer than e1, and it is padded with zeros when shorter than e1. When the DFs are the same, the lengths of e1 and e2 will be the same. Normalized correlation coefficients range from 0 (no correlation) to 1 (perfect correlation) and to  1 (perfect inverse correlation). Because e1 and e2 are not necessarily phase-aligned, the phase-optimal normalized cc was used for comparison, i.e. e 1 ð1 : wn Þ U e 2ϕ ð1 : wn Þ ccϕ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ½ðe 1 ð1 : wn Þ Ue 1 ð1 : wn ÞÞðe 2ϕ ð1 : wn Þ U e 2ϕ ð1 : wn Þ

ð6Þ

where ϕ denotes the phase of a2 that maximizes cc. To construct e2ϕ, the ensemble average e2 was adjusted by computationally wrapping it around the origin as needed so as to align with e1 for maximum correlation. Since the ensemble averages are used as periodic components to construct the basis vectors, their start and endpoints are arbitrary. Normalized correlation coefficients calculated using Eq. (6) were tabulated for all data, and separately for PV and FW. As a check on the sequence length for analysis, the measurements were repeated using the 1st and 2nd 4 s segment from the same data. All results were presented as mean 7standard

2129

deviation, and as coefficients of variation (standard deviation divided by mean). For comparison, the CFE-mean and the interval confidence level (ICL) parameters were calculated. The CFE-mean is defined as the average time duration between consecutive electrogram deflections during a specified time period [19]. We used similar parameter ranges to those suggested by the authors of that study: a refractory period (minimum) of 30 ms between counted deflections, absolute peak values within the range of 0.015–0.5 mV, a maximum deflection duration of 10 ms, and a time period of 8 s. The ICL was defined as the number of intervals between 50 ms and 120 ms for distinct electrogram deflections [20]. CFAE deflections were counted as distinct if their absolute peak values were within the range of 0.015–0.5 mV. This measurement was also done over 8 s intervals. Although it is a different kind of measurement, for completeness the DF was also calculated over each 8 s interval. 2.3. Statistical calculations Means of all data for paroxysmal versus persistent CFAE, and separately for PV and FW sites, were compared using the Mann– Whitney rank sum test (SigmaPlot 2004 for Windows Ver. 9.01, Systat Software, Chicago, and MedCalc ver. 9.5, 2008, MedCalc Software bvba, Mariakerke, Belgium). A value of p o0.05 was considered significant.

3. Results Examples of CFAE in persistent AF are shown in Fig. 1. The left superior, left inferior, right superior, and right inferior pulmonary vein recordings are given. In each trace there is little or no isoelectric interval, and the large deflections are time-varying. There is some periodicity evident in each of the signals, particularly in

0.2

0.10

LSPV

LIPV 0.1

0.05 0.0 0.00 -0.1 -0.05

-0.2

-0.3

-0.10 1

1

1000

1000

0.20

0.3

RSPV

RIPV

0.15

0.2 0.10

Magnitude

0.1

0.05

0.0

0.00

-0.1

-0.05

-0.2

-0.10 -0.15

-0.3

-0.20 1

1000

Sample Point

1000

1

Sample Point

Fig. 1. CFAE from the pulmonary vein ostia, longstanding persistent atrial fibrillation patient. LSPV – left superior pulmonary vein, LIPV – left inferior pulmonary vein, RSPV – right superior pulmonary vein, and RIPV – right inferior pulmonary vein.

2130

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

panels A, C and D. Note that only the first 1000 sample points,  1 s, are provided so that the electrogram detail can be observed. The 1st and 2nd 8 s DMs for these CFAE are shown in corresponding panels in Fig. 2. They are colored black and red, respectively, and have been adjusted for optimal phase alignment based on the crosscorrelation. When the DFs of the 1st versus 2nd 8 s segments are unequal, the lengths wn of the pair of vectors shown in each panel, given by Eq. (4), are also unequal (see panels B and D). There is evident similarity in the shape of the pairs, particularly for panels A, C and D, which have more periodic CFAE that can be observed in the corresponding panels of Fig. 1. The correlation of the DM pairs is given by the value of ccϕ at lower right in each panel of Fig. 2, as calculated using Eq. (6). Although there is not a great deal of overlap for the DM of Fig. 2C, there is a high degree of correlation. This is because the amplitudes and baseline levels of the traces are normalized by the denominator of Eq. (6). Since the shapes of the traces in panel C are otherwise quite similar, there is a large correlation. In contrast, normalization of the y-axis shift or scale would still not provide good overlap for the traces of panel B, and the ccϕ value is substantially lower, 0.378. Examples of CFAE in paroxysmal AF are provided in Fig. 3. As in Fig. 1, the left superior, left inferior, right superior, and right inferior pulmonary vein recordings are given. In each tracing there is no isoelectric segment and the large deflections change even more drastically over time as compared to the tracings for persistent AF in Fig. 1. There is some periodicity to the signals of panels B and C; however the large deflections change dramatically in shape from one cycle to the next. The corresponding 1st and 2nd 8 s ensemble averages at the DF for these CFAE are given in Fig. 4. There is partial overlap in the traces of Fig. 4, particularly for

40

LSPV

those with greater CFAE periodicity (see Fig. 3B and C). There is less correlation of the DM in most of the panels of Fig. 4 for the paroxysmal CFAE, as compared with those of persistent CFAE in Fig. 2. The periods of the DM for the paroxysmal data (Fig. 4) tend to be longer as compared to the persistent AF ensemble averages (Fig. 2), meaning that the DFs in paroxysmal AF tend to be lesser in frequency, as calculated using Eq. (4). Examples of power spectra are shown in Fig. 5. These spectra were all generated from CFAE acquired from the right superior pulmonary vein ostia. The top two panels, 1st and 2nd 8 s for persistent AF, were generated from the CFAE of Fig. 1C. The lower two panels, 1st and 2nd 8 s for paroxysmal AF, were generated from the CFAE of Fig. 3C. The persistent AF spectra appear to be less complex. The DF is evident at 6.7 Hz and is maintained in both of the top panels. There is a prominent subharmonic at  3.35 Hz in both panels A and B. The paroxysmal AF spectra, however, tend to be more complex. There is a higher and more complex spectral background level, as well as several prominent peaks, particularly in panel D. The DF shifts from  5.6 Hz in panel C to  5.4 Hz in panel D. The prominent peak at 11–11.2 Hz is the second harmonic. For the particular examples given in Figs. 1–5, which were representative of all persistent and paroxysmal CFAE data, there was more instability in paroxysmal AF frequency components. This tended to result in less correlation between DM pairs in paroxysmal AF, as the DMs were more changeable from 1st to 2nd 8 s segment (i.e., more time-varying). Classification based on the DM correlation coefficients is shown in Fig. 6. For means versus coefficients of variation, one paroxysmal and one persistent patient are classified incorrectly. Based

30

LIPV

20

20

10

0

0

-20

-10 -40 -20 -60

ccφ = 0.378

ccφ = 0.927 10

20

30

40

50

60

70

80

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

90 100 110 120 130 140

RSPV

40

30

RIPV

20

Magnitude

20

10 0

0

-10 -20

-20 -30

-40

-40 -60 ccφ = 0.901 10

20

30

40

50

60

70

80

90 100 110 120 130 140

Sample Point

ccφ = 0.810

-50 10

20

30

40

50

60

70

80

90 100 110 120 130 140 150

Sample Point

Fig. 2. Dominant morphology vectors for the CFAE of corresponding panels in Fig. 1. Black – 1st 8 s. Red – 2nd 8 s. ccϕ ¼normalized correlation coefficient with optimal phase alignment. The value of ccϕ ranges from 1 (perfect correlation) to 0 (no correlation). The relative amplitude and the y-axis shift of the traces does not change the value of ccϕ. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

0.2

2131

0.20

LSPV

LIPV

0.15

0.1 0.10 0.05

0.0

0.00 -0.1

-0.05 -0.10

-0.2

-0.15 -0.3

-0.20 1

1000

1

1000

0.2

RIPV

RSPV

0.3

0.1

Magnitude

0.2 0.0 0.1 -0.1 0.0 -0.2 -0.1 -0.3 -0.2 -0.4 1

1000

1000

1

Sample Point

Sample Point

Fig. 3. CFAE from the pulmonary vein ostia, paroxysmal atrial fibrillation patient. The same abbreviations as in Fig. 1.

LSPV

20

30 LIPV 20

10

10 0

0

-10 -10 -20 -20

-30 -40

ccφ = 0.480

-30 20

40

60

80

100

120

140

160

180

ccφ = 0.747 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160

200

RSPV

40

RIPV 20

30

Magnitude

10 20 0

10 0

-10

-10 -20 -20

ccφ = 0.837 10 20 30 40 50 60 70 80 90 100110120130140150160170180

Sample Point

ccφ = 0.292 10

20

30

40

50

60

70

80

90 100 110 120 130 140

Sample Point

Fig. 4. Dominant morphologic basis vectors for the CFAE of corresponding panels in Fig. 3. The same description as for Fig. 2.

2132

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

RSPV Persistent – 1st

4.0

Persistent 2nd 2.5

3.5 3.0

2.0 2.5 1.5

2.0 1.5

1.0 1.0 0.5

0.5 4

6

8

10

12

4

6

8

12

Paroxysmal – 2nd

Paroxysmal – 1st

2.0

10

1.6

1.8 1.4

Magnitude

1.6 1.2

1.4 1.2

1.0

1.0 0.8

0.8 0.6

0.6 4

6

8

10

12

4

6

Frequency (Hz)

8

10

12

Frequency (Hz)

Fig. 5. Frequency spectra for persistent CFAE: (A) 1st 8 s and (B) 2nd 8 s, and for paroxysmal CFAE: (C) 1st 8 s and (D) 2nd 8 s. It is evident from the paroxysmal AF spectra that there is less stability in the electrical activation pattern, as compared with persistent AF data.

Mean Values, Correlation Coefficients

0.6

0.5

0.4

0.3

0.2

0.1 0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

COV, Correlation Coefficients Fig. 6. Classification to distinguish persistent versus paroxysmal AF data. The linear discriminant function is shown as a straight dashed line.

on the linear discriminant function, COV alone would suffice for adequate classification. However, a nonlinear discriminant function could be useful to further improve classification accuracy. 3.1. Summary statistics The summary data for all CFAE is shown in Table 1 and follows the results shown in Figs. 1–5. The mean value of the normalized

correlation coefficient in both paroxysmal and persistent AF, as calculated using Eq. (6), is shown for all data, and separately for pulmonary vein and free wall data. In each case there is greater correlation between DM in persistent as compared with paroxysmal CFAE. Highly significant differences in the correlation coefficients of DMs for persistent, versus the correlation coefficients for paroxysmal AF, are present for all data combined and for pulmonary vein data.

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

Table 1 Normalized correlation coefficients for dominant morphology – pooled data. Data

Persistent AF

Paroxysmal AF

Significance

All PV FW

0.6197 0.219 0.6167 0.211 0.625 7 0.238

0.502 7 0.190 0.4617 0.181 0.582 7 0.186

p o 0.001 p o 0.001 p ¼ 0.345

All – pooled data from all locations. PV – data from the ostia of the pulmonary veins only. FW – data from the left atrial free wall only.

Table 2 Normalized correlation coefficients for dominant morphology by location. Location

Persistent

Paroxysmal

Significance

LSPV LIPV RSPV RIPV ANT POS

0.6637 0.202 0.5577 0.187 0.654 7 0.201 0.5917 0.247 0.6747 0.220 0.5757 0.250

0.4147 0.153 0.454 7 0.179 0.5007 0.185 0.4787 0.209 0.6477 0.163 0.5187 0.191

p o 0.001 p ¼ 0.155 p ¼ 0.048 p ¼ 0.127 p ¼ 0.579 p ¼ 0.425

LSPV – left superior pulmonary vein, LIPV – left inferior pulmonary vein, RSPV – right superior pulmonary vein, RIPV – right inferior pulmonary vein, ANT – anterior left atrial free wall, and POS – posterior left atrial free wall.

Table 3 Normalized correlation coefficients for individual patients – all data. Type

Persistent AF

Paroxysmal AF

Patient

Mean

COV

Mean

COV

1 2 3 4 5 6 7 8 9 10 MN SD

0.707 0.577 0.564 0.705 0.587 0.699 0.561 0.606 0.510 0.618 0.613nn 0.068

0.298 0.180 0.311 0.298 0.574 0.256 0.290 0.434 0.402 0.405 0.345† 0.111

0.495 0.475 0.537 0.454 0.529 0.471 0.672 0.451 0.483 – 0.507nn 0.069

0.432 0.393 0.308 0.387 0.325 0.378 0.454 0.338 0.394 – 0.379† 0.048

COV – coefficient of variation. nn

†

p ¼0.004. p ¼not significant.

The normalized correlation coefficients at individual anatomic locations are shown in Table 2. At each location, the DM from 1st and 2nd 8 s sequences of each CFAE is more correlated for persistent AF data. The significance of the difference in mean correlation values is given by the p value. There is a significant difference at the left superior pulmonary vein (po 0.001) and at the right superior pulmonary vein (p o0.05). The values of normalized correlation coefficients by patient are shown in Table 3. The mean values averaged for persistent and paroxysmal AF correspond to the mean values in Table 1, all data, except for rounding. The coefficients of variation are also provided. There is more spatial variability in the correlation between DM for paroxysmal AF patients (average coefficient of variation of 0.379) as compared to persistent AF patients (average coefficient of variation of 0.345) but this difference did not rise to the level of significance. In Table 4 are shown the results for 4 s data. As for the 8 s comparisons of Table 3, mean DM correlation coefficients are greater for persistent as compared with paroxysmal AF data. However there were no significant differences, suggesting that

2133

Table 4 Normalized correlation coefficients for individual patients – all data 4096 points (4 s). Type

Persistent AF

Paroxysmal AF

Patient

Mean

COV

Mean

COV

1 2 3 4 5 6 7 8 9 10 MN SD

0.622 0.498 0.581 0.676 0.430 0.648 0.601 0.671 0.553 0.586 0.587 0.077

0.371 0.283 0.303 0.374 0.751 0.313 0.214 0.339 0.148 0.327 0.343 0.160

0.511 0.513 0.482 0.537 0.581 0.540 0.508 0.473 0.724 – 0.541 0.076

0.540 0.355 0.405 0.322 0.164 0.462 0.247 0.154 0.103 – 0.306 0.150

COV – coefficient of variation. p ¼not significant.

Table 5 Type

Persistent AF

Paroxysmal AF

Significance

91.23 7 19.65 90.177 19.00 95.62 7 42.30 92.30 7 22.44 5.54 7 1.22 5.69 7 1.12

p ¼0.255 p ¼0.027 p ¼0.237 p ¼0.043 p o0.001 p o0.001

(B) CFE-mean, ICL, and DF (two 8 s intervals) ICL 94.54 7 18.52 90.707 19.28 CFE-M 89.017 20.99 93.96 7 33.81 DF 6.217 1.01 5.617 1.17

p ¼0.020 p ¼0.025 p o0.001

(A) CFE-mean ICL, and DF (8 s intervals) ICL – 1st 8 s 94.247 18.48 ICL – 2nd 8 s 94.85 7 18.65 CFE-M 1st 8 s 89.147 20.39 CFE-M 2nd 8 s 88.89 7 21.66 DF – 1st 8 s 6.20 7 1.00 DF – 2nd 8 s 6.22 7 1.03

ICL in ♯ of deflections per 8 s. CFE-M in milliseconds.

8 s data provides better results. In Table 5 are shown results using CFE-mean and ICL. For 1st versus 2nd 8 s, there were moderately significant differences in persistents versus paroxysmals for the 2nd 8 s segment, ICL and CFE-mean parameters (p o0.05), as shown in Table 5A. When the data from both 8 s segments were combined, there were again moderate significance differences between persistents and paroxysmals, ICL and CFE-mean parameters (p o0.05), as shown in Table 5B. The DF parameter is highly significant for both 8 s intervals separately and for the combined intervals (Table 5; p o0.001). This measurement accounts for differences in the value of a single frequency, as compared with the DM which is indicative of changes in fractionated electrogram shape, and therefore changes in the pattern of electrical activation.

4. Discussion 4.1. Summary In this study the concept of a dominant morphology (DM) was introduced for CFAE signals. The DM is defined as the ensemble average of signal segments at the dominant frequency. The DM is representative of the basic shape of the CFAE from one electrical activation interval to the next. The DM may appear more or less like the original CFAE, depending on the periodicity and regularity of the CFAE deflections, and the degree to which the DF dominates the power spectrum. DMs were compared for 1st versus 2nd 8 s sequences of CFAE recordings as an estimate of the stability of the

2134

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

time and frequency components. It was found at all individual anatomic locations and for the left atrium as a whole, that there is greater correlation between the two segments in persistent AF. The greater correlation means that the DM is more stable over time in persistent as compared to paroxysmal CFAE. The difference in correlation was highly significant for left superior pulmonary vein recordings and moderately significant for the right superior pulmonary vein recordings. It was also found that the spatial variation in the correlation of DM pairs, in terms of the COV, is greater in paroxysmal than in persistent CFAE data, although these values did not rise to the level of significance (Tables 3 and 4). The greater spatial variation in paroxysmal AF suggests that there may be a less centralized and consistent source driving AF in these patients. This is in agreement with a recent study in which the DF in paroxysmal AF was found to be spatiotemporally unstable as compared with persistent AF, and is therefore not indicative of fixed drivers [21]. 4.2. Clinical correlates In this study the concept of the DM was introduced and tested. The DM adds to the arsenal of descriptors that have been developed to characterize CFAE, which now include the dominant amplitude (DA), and the mean and standard deviation in spectral profile (MP and SP). The DM is a combined morphologic and frequency descriptor. The use of DM and DA provide information about the morphologic detail at the DF. These parameters go beyond the DF measurement, by characterizing CFAE shape according to the morphology of the main frequency component, rather than measuring frequency itself. The DA but not the DM can be extracted using the Fourier transform. The Fourier basis is a general basis consisting of sinusoids. Each Fourier frequency component, a sinusoid, has an amplitude and frequency, but it does not have an associated morphology. The DM extracted as described in Section 2 using the ensemble averaging method, provides the morphology of the CFAE at the DF. It represents the main periodic shape of the signal, the first statistical moment, which is sometimes but not always identifiable in the original CFAE data (Figs. 1–4). By knowing the spatiotemporal variability in the DM, as was calculated in this study, it is possible to infer the stability of drivers of electrical activation in the vicinity of the recording electrode. It would be expected that stable drivers would have spatiotemporal stability in the DM, that is, a high degree of correlation from 1st to 2nd 8 s recordings, and a similarly high degree of correlation at spatially distinct recording locations. A site with highest degree of temporal correlation would be proposed to be a candidate catheter ablation site, since the stability of such a site would likely be indicative of the presence of a stable driver in the vicinity of the recording electrode. Since such sites appear to exist more commonly in persistent AF, it is possible that ablation at a subset of CFAE recording sites in these patients would do as well to eliminate AF as compared with ablation at all CFAE recording sites. Such a constraint may be helpful to reduce morbidity due to the procedure. Areas where DM is stable that also have high DF may be of particular interest for catheter ablation [22]. Although not investigated in the current study, other ensemble average vectors besides the DM could be compared to provide additional information. For example it might be useful to compare the morphology of ensemble averages at other tall peaks in the frequency spectrum, which may be indicative of secondary, independent drivers of electrical activity. Although comparisons were made from 1st 8 s to 2nd 8 s and from 1st 4 s to 2nd 4 s of a 16 s time series in this study, comparisons could also be made over longer or shorter time intervals, as well as between segments disparate in time. For example the long-term stability of the DM

could be estimated by extracting 8 s segments from CFAE that are separated by 1 min or more in time. 4.3. Limitations The study was done using CFAE recordings from a limited number of sites and a limited number of patients. The investigation should be repeated prospectively with larger numbers of recording sites and patients. DM were defined as the ensemble average at the DF. The DF often changed from 1st to 2nd 8 s segments. Thus for simplicity, DF ensemble averages which could have differing vector lengths were compared. Although comparisons of ensemble averages at the same frequency may have some relevance (for example use of the DF of the 1st 8 s segment to also extract the ensemble average of the 2nd 8 s segment for comparison), slight temporal shifts in DF over 16 s are common. Hence the power at the frequency of the DF in one 8 s segment may shift to low levels at that same frequency in the other 8 s segment, as for example in Fig. 5C versus D, which would likely render such a comparison less relevant.

5. Conclusions The DM can be extracted and compared in paroxysmal and persistent CFAE. There is higher temporal variability in the DM of paroxysmal CFAE. There is also higher spatial variability in the DM of paroxysmal CFAE, although this does not rise to the level of significance (Tables 3 and 4). Extraction and analysis of the DM show that it is useful to quantitate electrogram morphology at the DF so that CFAE in paroxysmal versus persistent AF patients can be compared. The greater spatiotemporal variability in paroxysmal AF is suggestive of more instability in the electrical activity of these patients, and is in agreement with prior studies [13–16]. The greater spatiotemporal correlation in persistent CFAE is suggestive of the presence of more stable, intransigent drivers of AF in these patients, perhaps due to intractable structural remodeling. The DM, as well as ensemble averages arising from secondary peaks in the frequency spectra, may be indicative of the morphology and characteristics of AF drivers of electrical activity that should be sought for catheter ablation.

Conflict of interest statement The authors have no conflicts of interest. References [1] C. Pappone, S. Rosanio, G. Oreto, M. Tocchi, F. Gugliotta, G. Vicedomini, A. Salvati, C. Dicandia, P. Mazzone, V. Santinelli, S. Gulletta, S. Chierchia, Circumferential radiofrequency ablation of pulmonary vein ostia: a new anatomic approach for curing atrial fibrillation, Circulation 102 (2000) 2619–2628. [2] C. Pappone, S. Rosanio, G. Augello, G. Gallus, G. Vicedomini, P. Mazzone, S. Gulletta, F. Gugliotta, A. Pappone, V. Santinelli, V. Tortoriello, S. Sala, A. Zangrillo, G. Crescenzi, S. Benussi, O. Alfieri, Mortality, morbidity, and quality of life after circumferential pulmonary vein ablation for atrial fibrillation: outcomes from a controlled nonrandomized long-term study, Journal of the American College of Cardiology 42 (2003) 185–197. [3] H. Oral, B.P. Knight, H. Tada, M. Ozaydin, A. Chugh, S. Hassan, C. Scharf, S.W. Lai, R. Greenstein, F. Pelosi Jr, S.A. Strickberger, F. Morady, Pulmonary vein isolation for paroxysmal and persistent atrial fibrillation, Circulation 105 (2002) 1077–1081. [4] P. Sanders, O. Berenfeld, M. Hocini, P. Jaïs, R. Vaidyanathan, L.F. Hsu, S. Garrigue, Y. Takahashi, M. Rotter, F. Sacher, C. Scavée, R. Ploutz-Snyder, J. Jalife, M. Haïssaguerre, Spectral analysis identifies sites of high-frequency activity maintaining atrial fibrillation in humans, Circulation 112 (2005) 789–797. [5] S. Willems, H. Klemm, T. Rostock, B. Brandstrup, R. Ventura, D. Steven, T. Risius, B. Lutomsky, T. Meinertz, Substrate modification combined with pulmonary

E.J. Ciaccio et al. / Computers in Biology and Medicine 43 (2013) 2127–2135

[6]

[7]

[8]

[9]

[10]

[11] [12]

[13]

[14]

vein isolation improves outcome of catheter ablation in patients with persistent atrial fibrillation: a prospective randomized comparison, European Heart Journal 27 (2006) 2871–2878. S. Dixit, F.E. Marchlinski, D. Lin, D.J. Callans, R. Bala, M.P. Riley, F.C. Garcia, M.D. Hutchinson, S.J. Ratcliffe, J.M. Cooper, R.J. Verdino, V.V. Patel, E.S. Zado, N.R. Cash, T. Killian, T.T. Tomson, E.P. Gerstenfeld, Randomized ablation strategies for the treatment of persistent atrial fibrillation: RASTA study, Circulation: Arrhythmia and Electrophysiology 5 (2012) 287–294. K. Nademanee, J. McKenzie, E. Kosar, M. Schwab, B. Sunsaneewitayakul, T. Vasavakul, C. Khunnawat, T. Ngarmukos, A new approach for catheter ablation of atrial fibrillation: mapping of the electrophysiologic substrate, Journal of the American College of Cardiology 43 (2004) 2044–2053. M. Holm, S. Pehrson, M. Ingemansson, L. Sörnmo, R. Johansson, L. Sandhall, M. Sunemark, B. Smideberg, C. Olsson, S.B. Olsson, Non-invasive assessment of the atrial cycle length during atrial fibrillation in man: introducing, validating and illustrating a new ECG method, Cardiovascular Research 38 (1998) 69–81. O. Berenfeld, R. Mandapati, S. Dixit, A.C. Skanes, J. Chen, M. Mansour, J. Jalife, Spatially distributed dominant excitation frequencies reveal hidden organization in atrial fibrillation in the Langendorff-perfused sheep heart, Journal of Cardiovascular Electrophysiology 11 (2000) 869–879. G.W. Botteron, J.M. Smith, A technique for measurement of the extent of spatial organization of atrial activation during atrial fibrillation in the intact human heart, IEEE Transactions on Biomedical Engineering 42 (1995) 579–586. G.W. Botteron, J.M. Smith, Quantitative assessment of the spatial organization of atrial fibrillation in the intact human heart, Circulation 93 (1996) 513–518. R. Alcaraz, F. Sandberg, L. Sörnmo, J.J. Rieta, Application of frequency and sample entropy to discriminate long-term recordings of paroxysmal and persistent atrial fibrillation, in: Conference Proceedings of the IEEE Engineering in Medicine and Biology Society, 2010, vol. 2010, pp. 4558–4561. E.J. Ciaccio, A.B. Biviano, W. Whang, A. Gambhir, H. Garan, Spectral profiles of complex fractionated atrial electrograms are different in longstanding and acute onset atrial fibrillation, Journal of Cardiovascular Electrophysiology 23 (2012) 971–979. E.J. Ciaccio, A.B. Biviano, W. Whang, A. Gambhir, H. Garan, Different characteristics of complex fractionated atrial electrograms in acute paroxysmal versus long-standing persistent atrial fibrillation, Heart Rhythm 7 (2010) 1207–1215.

2135

[15] E.J. Ciaccio, A.B. Biviano, W. Whang, J.A. Vest, A. Gambhir, A.J. Einstein, H. Garan, Differences in repeating patterns of complex fractionated left atrial electrograms in longstanding persistent atrial fibrillation as compared with paroxysmal atrial fibrillation, Circulation: Arrhythmia and Electrophysiology 4 (2011) 470–477. [16] E.J. Ciaccio, A.B. Biviano, W. Whang, A.L. Wit, J. Coromilas, H. Garan, Optimized measurement of activation rate at left atrial sites with complex fractionated electrograms during atrial fibrillation, Journal of Cardiovascular Electrophysiology 21 (2010) 133–143. [17] E.J. Ciaccio, A.B. Biviano, W. Whang, J. Coromilas, H. Garan, A new transform for the analysis of complex fractionated atrial electrograms, BioMedical Engineering OnLine 10 (2011) 35. [18] E.J. Ciaccio, A.B. Biviano, W. Whang, A. Gambhir, H. Garan, Improved frequency resolution for characterization of complex fractionated atrial electrograms, BioMedical Engineering OnLine 11 (2012) 17. [19] M.K. Stiles, A.G. Brooks, B. John, Shashidhar, L. Wilson, P. Kuklik, H. Dimitri, D.H. Lau, R.L. Roberts-Thomson, L. Mackenzie, S. Willoughby, G.D. Young, P. Sanders, The effect of electrogram duration on quantification of complex fractionated atrial electrograms and dominant frequency, Journal of Cardiovascular Electrophysiology 19 (2008) 252–258. [20] D. Scherr, D. Dalal, A. Cheema, S. Nazarian, I. Almasry, K. Bilchick, A. Cheng, C.A. Henrikson, D. Spragg, J.E. Marine, R.D. Berger, H. Calkins, J. Dong, Longand short-term temporal stability of complex fractionated atrial electrograms in human left atrium during atrial fibrillation, Journal of Cardiovascular Electrophysiology 20 (2009) 13–21. [21] J.W.E. Jarman, T. Wong, P. Kojodjojo, H. Spohr, J.E. Davies, M. Roughton, D.P. Francis, P. Kanagaratnam, V. Markides, D.W. Davies, N.S. Peters, Spatiotemporal behaviour of high dominant frequency during paroxysmal and persistent atrial fibrillation in the human left atrium, Circulation: Arrhythmia and Electrophysiology 5 (2012) 650–658. [22] P. Sanders, O. Berenfeld, M. Hocini, P. Jais, R. Vaidyanathan, L.F. Hsu, S. Garrigue, Y. Takahashi, M. Rotter, F. Sacher, C. Scavee, R. Ploutz-Snyder, J. Jalife, M. Haissaguerre, Spectral analysis identifies sites of high-frequency activity maintaining atrial fibrillation in humans, Circulation 112 (2005) 789–797.