cROSS CORRELATION ANALYSIS FOR THE ESTIMATION OF DEFIBRILLATION SHOCK ENERGY

Z. F. Syed¹,E. Vigmond¹, S. Kimber², and L.J. Leon¹
¹Dept. Of Electrical and Computer Engineering University of Calgary,
2500 University Drive NW. Calgary, Alberta, and ²Dept. of Medicine University of Alberta, Edmonton Alberta

Abstract: Electrograms recorded from 15 patients during the implantation of Implantable Cardioverter Defibrillators (ICD) were analyzed. We found that there was a slightly lower variation in the maximum beat to beat cross correlation, and a lower frequency associated with successful 6 joule defibrillation shocks than for unsuccessful shocks.

INTRODUCTION

Although the modern ICD has seen tremendous change since it first appeared [1], there are still many avenues for improvement. The most important of these are 1) the reduction of energy requirements, and 2) the increase in the safety-factor for defibrillation. At present, defibrillation shock strengths are pre-programmed at implantation. This makes the fundamental assumption that all fibrillation in a given patient is identical and consequently all episodes would require the same shock strength to terminate them. One possible approach to reduce the required shock strength is to find an “intelligent approach to the choice of shock strength”. We have begun to investigate the possibility of choosing the defibrillation shock strength based upon the form and dynamics of the electrical activity observed by the sensing electrode.

Using optical mapping techniques Witkowski et al. [2] reference missing showed that VF begins as a single rotor which breaks down into a number of distinct meandering waves. They called this initial relatively well organized phase acute VF. It was characterized by the fact that there were a limited number (3 or less) of easily identifiable spiral waves present, and the maximum cross-correlation over space and time was relatively high (>0.6). Interestingly, several studies have shown that the probability of success of defibrillation correlates directly with the level of organization of the electrical activity of the ventricular fibrillation exhibits [3,4,5].

In this work we have begun to examine if “organizational parameters” such as maximum cross-correlation coefficient and its standard deviation, and/or mean frequency might be of some value in predicting required shock strength. We used the cross correlation function along with the frequency analysis from 15 different patient records and compared these measures to the success of a 6 joule defibrillation shock.

METHODS

Sensing electrode data was collected from 15 patients undergoing implantation of Medtronic ICD's as part of a study evaluating the Medtronic Model 6944 lead. Each patient record contained two electrogram signals. An atrial bipolar electrogram (tip to ring), and a ventricular bipolar electrogram(also tip to ring). The Ventricular electrograms were analysed in the following manner: 1) Raw signals were differentiated. 2) A 3 point box-car filter was used to filter the differentiated signal. 3) Beats were identified using the maximum negative slope in the differentiated signal. 4) Mean frequency (beats/sample) was calculated for the window from onset of sampling until the shock was actually delivered. 5) Finally the maximum cross correlation function of the differentiated signal was calculated. The differentiated signals were used to eliminate baseline drift.

The cross correlation statistic at time t for a lag and a window of length n is defined as

The maximum cross correlation (max CC) was taken as the maximum of when was allowed to vary over the four subsequent beats.

RESULTS

As mentioned above, the data was taken from patients undergoing implant. At which time an up-down protocol is used to determine the initial shock strength. In this preliminary study three measures of organization were calculated for a series of episodes of VF in each of 15 patients. Each of these episodes was treated with a 6 joule shock. The organization measures used were the mean max CC, the mean standard deviation of the max CC, and the frequency.

Figure 1 shows results from two episodes in the same patient. Within each panel the top trace shows the electrogram waveform, the middle trace, the derivative of the signal, and the bottom trace the maximum cross-correlation statistic. The top panel shows results from an unsuccessful shock while the bottom panel shows results from a successful shock. Although both of these episodes seem quite similar on the surface there are subtle differences between them. In particular for the successful defibrillation the mean max CC was higher, 0.68 as opposed to 0.65, the mean standard deviation was lower, 0.25 vs 0.32. It is also important to note the trend in CC statistics. In the case of the failed defibrillation shock the max CC curve is decreasing, suggesting a decrease in the level of organization.

As a first attempt at analyzing the data we divided the data into two groups, 6 successful defibrillation shock episodes, and 9 unsuccessful shocks. We found that the mean max CC was very slightly higher for the successful shocks 0.72 vs 0.71 for the unsuccessful shocks. There was a more significant difference in the standard deviation of the max CC statistic, it had a mean of 0.32 for the successful shocks and 0.36 for the unsuccessful shocks. The most significant difference between the two was the frequency, 4.7 beats/sec for successful shocks and 5.3 beats/second for unsuccessful shocks.

Figure 1 Top Panel: Unsuccessful 6 joule shock top trace electrogram waveform, middle trace, the derivative of the signal, and the bottom trace the maximum cross-correlation statistic. Bottom panel: Successful 6 joule shock in the same patient, same layout as Top Panel.

DISCUSSION

The goal of this preliminary study was to investigate the possible predictors of shock strength required to defibrillate. We have developed a number of software tools to be used in this work. In subsequent work we will compare these statistics for successful and unsuccessful shocks of the same shock strength in the same patient. In doing so we hope to find predictors of defibrillation threshold.

Acknowledgments: We are grateful to Medtronic Corporation for supplying data for the study. This work was supported by the NSERC, and startup grants from the University of Calgary.

REFERENCES

[1] Mirowski M, Reid PR, Mower MM, Watkins L, Gott JA, Schauble JF, Langer A, Heilman MS, Kolenik SA, Fischell RE, Weisfeldt ML: Termination of malignant ventricular arrhythmias with an implanted automatic defibrillator in human beings. N Engl J Med 1980, 303

[2] Spatiotemporal evolution of ventricular fibrillation Frank Witkowski, L. Joshua Leon, Patricia A. Penkoske, Wayne R. Giles, Mark L. Spano, William L. Ditto & Arthur Winfree Nature Vol 392:78-82 (1998)

[3] Karagueuzian HS and Chen P. Fibrillation and Defibrillation: The Odd Couple?. Journal of Cardiovascular Electrophysiology, June 2000; Vol. 11, 642-644

[4] Hsia PW, Fendelander L, Harrington G. et al. Defibrillation Success Is Associated With Myocardial Organization: Spatial Coherence as a New Method of Quantifying the Electrical Organization of the Heart. Journal of Electrocardiology. 1996; 29: 189-197

[5] Fendelander L, Hsia PW and Damiano RJ. Spatial Coherence: A New Method of Quantifying Myocardial Electrical Organization Using Multichannel Epicardial Electrograms. Journal of Electrocardiology. January 1997; 30(1): 9 – 19

[6] Brown CG, Dzwonczyk R. “Signal analysis of the human ECG during VF – frequency and amplitude parameters as predictors of successful countershock. Annals of Emergency Medicine 1996; Vol. 27 (Number 2):184-88

[7] Strohmenger H, Lindner KH and Brown CG. Analysis of the Ventricular Fibrillation ECG Signal Amplitude and Frequency Parameters as Predictors of Countershock Success in Humans. Chest 1997; 111:584-89

[[8] Karagueuzian HS and Chen P. Fibrillation and Defibrillation: The Odd Couple?. Journal of Cardiovascular Electrophysiology, June 2000; Vol. 11, 642-644