Body-Surface Potential Maps Can Identify
Substrate for Ventricular Arrhythmias

Milan B. Horácek

Department of Physiology & Biophysics, Department of Medicine (Cardiology)
Dalhousie University, Halifax, Nova Scotia, Canada

Correspondence: Milan B Horácek, 4-P2 Sir Charles Tupper Med. Bldg., Dalhousie University, Halifax, NS, Canada B3H 4H7.
E-mail: milan.horacek@dal.ca, phone 902-494-1306, fax 902-494-1685

1.  Introduction

Prior myocardial infarction (MI) is the most common cause of sustained ventricular tachycardia (VT).^1,2 Options for management of malignant VT include implantable cardioverter defibrillators (ICDs),³ antiarrhythmic drugs, arrhythmia surgery, and catheter ablation.^1,2,4-9 Radio-frequency catheter ablation has largely supplanted surgical therapy,⁴ and is an alternative and/or an adjunct to pharmacological or ICD therapy;⁵ however, radiofrequency ablation of VT due to a prior MI is still one of the most challenging procedures in clinical electrophysiology.⁶ Successful ablation of scar-related VT requires an understanding of the VT’s mechanism within the underlying electroanatomic substrate. This can be achieved by percutaneous endocardial mapping with a single catheter, which is steered to multiple endocardial sites^{1, 7} and creating an endocardial map during sinus rhythm. This alone, however, is not sufficient to predict successful ablation sites; some tachycardias are “unmappable” with a point-by-point mapping and alternative methods for identifying VT's exit sites during sinus rhythm or pacing have to be sought.⁸ These techniques include limited entrainment mapping⁶ and pacemapping in the peri-infarct border zone.⁹ They can be further assisted by a simultaneous body-surface mapping during VT or pacing; this will, no doubt, happen in the near future.

Ventricular tachyarrhythmias associated with MI occur in two stages. During the acute phase of MI, polymorphic VT that can degenerate into ventricular fibrillation is most common. During the healing phase of MI, the infarction zone undergoes structural changes; fibrosis creates areas of conduction block and slows conduction through myocyte pathways by increasing separation of myocyte bundles in the infarct’s border zone.¹⁰ These pathways can support stable reentry circuits, giving rise to monomorphic VTs, when an appropriate trigger occurs. With present management of acute MI, fewer than 5% of acute MI survivors have inducible VT early after the acute event.¹¹ Patients with large infarcts—often those who were not successfully reperfused—are at greatest risk for VT. Reentrant VTs often exit from subendocardial sites of infarcted myocardium adjacent to the densely scarred tissue; viable bundles of myocardial fibers become embedded in extensive fibrous tissue.¹⁰ Accurate determination of an electroanatomic substrate for arrhythmia (slow conduction, conduction block, and isthmuses) in patients with VT is essential for guiding catheter ablation;⁴ this task is now facilitated by nonfluoroscopic electroanatomic mapping.¹²

The theme of this session is to demonstrate the role of body-surface potential mapping in identification of the arrhythmogenic substrate^13-16 and in guiding the ablation of arrhythmogenic regions in patients with VTs.^17-20 In the diagnostic workup for catheter ablation, and during ablation itself, body-surface potential mapping can serve both as a first-step localization method, and as an on-line method that can supplement the pacemapping and the endocardial electroanatomic mapping.¹² The focus of this paper is on applications of statistical methods that we have been using to identify diagnostic features of body-surface potential maps that are associated with arrhythmogenicity.

2.  Methods

The study sample consisted of 102 VT patients (70 with postinfarction substrate and 32 with other etiology) and 102 patients with remote MI who had no history of ventricular arrhythmias. Electrophysiologic studies confirmed the inducibility of VT in all but 9% of patients in the former group. This population is described in detail elsewhere.¹⁶ All patients underwent body-surface potential mapping, using 117 unipolar thoracic leads recorded with reference to the Wilson central terminal. From the 15-s recordings, averaging yielded a single complex for each lead; the QRS onset (R_on) and T-wave offset (T_off) were determined and QRST areas (in μVs) were calculated as time-integrals of ECG signal from R_on to T_off. The distributions of these values were displayed as isocontour maps.^{15, 16} The Karhunen-Loeve (K-L) transform was used to reduce the dimensionality of the input data consisting of an ensemble, x, of n m-dimensional random vectors (n=204; m=117), each representing the set of QRST-integral values for one patient. An eigenvalue and eigenvector analysis of the sample covariance matrix, C, yielded a square matrix, T, of m eigenvectors and a diagonal matrix, Λ, of eigenvalues, such that

where T^T denotes a transpose of T.²³ For each subject, denoted by subscript i, the K-L transform was defined by a relationship²²

which assigned an output vector of K-L coefficients, y_i, to an input vector of QRST-integral measurements x_i. We truncated vectors of K-L coefficients to k terms (k<<m) and then reconstructed the distributions of QRST integrals for each subject by a reverse transform. The 16 eigenvectors derived from the sample covariance matrix were plotted as eigenmaps,¹⁶ which represent the principal patterns of the QRST-integral distributions on the body surface. To select features that contained the diagnostic information for classifying the two constituent groups, subsets of K-L coefficients that contain the diagnostic information for separating the groups were selected by a stepwise discriminant analysis, using both forward and backward selection. Linear discriminant functions were calculated²⁴ for each set of selected features, and the classification performance that can be expected on a prospective population for different subsets of features was estimated by the bootstrap method.²⁵ Using the bootstrap without replacement, one half of the cases of each group was randomly assigned to a training set and the rest to a test set; the linear discriminant function was then computed for each training set and applied to the test set. Sensitivity, specificity, diagnostic performance (DP), positive predictive value, and negative predictive value were then calculated; the latter two measures depend on the prevalence of VT in the total patient sample. This procedure was repeated 1000 times and the resulting data provided estimates of classification performance, expressed as mean ± SD for all 1000 training and test sets. Finally, we checked how the results of our QRST-integral analysis correlate with the other clinical variables by calculating the correlation matrix, with associated p values, for all K-L coefficients and the selected clinical variables.

3.  Results

We first reduced the entire pattern space, consisting of 117 QRST-integral values for all patients, to 16 principal patterns (eigenvector maps),¹⁶ and then examined how these patterns contribute to QRST-integral maps in each diagnostic group. The stepwise linear discriminant analysis showed that the spatial features with the best ability to discriminate between two groups were drawn (in that order) from eigenvector maps 6, 4, 13, 5, 1, 2, 11, 14, 10, 15, and 9, for both forward and backward selection. The bootstrap method was used to calculate the means and standard deviations of classification indices for the increasing number of features (added in the order selected by the stepwise linear discriminant analysis) for the patient population consisting of the entire VT group (n = 102) and Non-VT group (n = 102); randomly selected training and test sets had 102 patients each, with both groups equally represented. The results showed that the percentage of correctly classified patients (DP) for both the training sets and the test sets increased monotonically as the number of features increased from 1 to 8. Further increase in the number of features increased the percentage of correctly classified patients for the training sets, but there was a deterioration of DP for the test sets. For a set of eight features, the classifier's sensitivity and specificity for detecting patients with VT in the test sets were (mean ± SD) 90.3 ± 4.3% and 78.0 ± 6.1%, respectively. The same procedure was then applied to the subpopulation of 172 postinfarction patients with VT (n = 70) and without VT (n = 102). The results (Fig. 1) were similar as for the entire VT group; the mean DP of the classifier for both training and test sets (n was 86 in each, with the VT group and Non-VT group represented by 35 and 51 patients, respectively) increased monotonically till the number of features reached 10, and then it declined. For an optimal subset of ten features (y₆, y₄, y₂, y₁, y₁₃, y₅, y₁₅, y₁₁, y₁₄, and y₉), the classifier's sensitivity and specificity for detecting postinfarction patients prone to VT in test sets were 90.0 ± 5.2% and 82.4 ± 5.3%, respectively.

4.  Discussion

The spatial features of the body-surface potential distributions during both the depolarization and the repolarization of ventricular myocardium are likely to reflect an anatomic and electrophysiologic substrate for sustained ventricular arrhythmias. We chose to assess the spatial distribution of ventricular primary repolarization properties, whose disparities are known to be associated with arrhythmogenesis.

The results of our studies indicate that spatial features extracted from QRST-integral maps provide diagnostic information from which a patient's vulnerability to VT can be predicted. The appropriately weighted combination of these features appears to reflect accurately the substrate for ventricular arrhythmias in members of the two distinct diagnostic groups. We have therefore provided empirical statistical evidence linking alterations in primary repolarization properties, measured from electrocardiographic QRST-integral maps, and the risk for sustained ventricular arrhythmias. Based on estimates obtained by the bootstrap procedure, the sensitivity for detecting patients at risk for VT can be expected to be around 90%, and it should be accompanied by the specificity of about 80%. These results compare favorably with the diagnostic performance achieved by electrophysiologic testing, but, of course, unlike the latter method, body-surface mapping is noninvasive.

Acknowledgements

The work presented here was supported by grants from the Canadian Institutes of Health Research and from the Heart & Stroke Foundations of Nova Scotia and Alberta. The clinical studies were carried out in the Queen Elizabeth II Health Sciences Centre in Halifax and in the Foothills Hospital in Calgary; they were performed in collaboration with cardiologists Drs. Martin J. Gardner, Terrence J. Montague, L. Brent Mitchell, and Eldon R. Smith. My special thanks go to my former students Drs. Cheryl Hubley-Kozey, Rok Hren and Cindy J. Penney, and to the members of the group in the Department of Physiology & Biophysics, Brian K. Hoyt, Paul MacInnis, James W. Warren, and Robert Potter for their contributions to the body-surface potential mapping projects carried out in our institution.

References

1. Delacretaz E and Stevenson WG. Catheter ablation of VT in patients with coronary heart disease. Part I: Mapping. Pacing Clin Electrophysiol 2001;24:1261-77.

2. Delacretaz E and Stevenson WG. Catheter ablation of VT in patients with coronary heart disease. Part II: Clinical aspects, limitations, and recent developments. Pacing Clin Electrophysiol 2001;24:1403-11.

3. Moss AJ, Zareba W, Hall WJ, Klein H, Wilber DJ, Cannom DS, Daubert JP, Higgins SL, Brown MW, and Andrews ML. Prophylactic implantation of a defibrillator in patients with myocardial infarction and reduced ejection fraction. N Engl J Med 2002;346:877--83.

4. Stevenson WG. VT after myocardial infarction: From arrhythmia surgery to catheter ablation. J Cardiovasc Electrophysiol 1995;6:942-50.

5. Stevenson WG, Friedman PL, and Sweeney MO. Catheter ablation as an adjunct to ICD therapy (editorial). Circulation 1997;96:1378-80.

6. Soejima K, Suzuki M, Maisel WH, Brunckhorst CB, Delacretaz E, Blier L, Tung S, Khan H, and Stevenson WG. Catheter ablation in patients with multiple and unstable VTs after myocardial infarction. Circulation 2001;104:664-9.

7. Hsia HH and Marchlinski FE. Characterization of the electroanatomic substrate for monomorphic VT in patients with nonischemic cardiomyopathy. Pacing Clin Electrophysiol 2002;25:1114-27.

8. Brunckhorst CB, Stevenson WG, Jackman WM, Kuck KH, Soejima K, Nakagawa H, Cappato R, and Ben-Haim SA. Ventricular mapping during atrial and ventricular pacing. Relationship of multipotential electrograms to VT reentry circuits after myocardial infarction. Eur Heart J 2002;23:1131--8.

9. Wilber DJ, Kopp DE, Glascock DN, Kinder CA, and Kall JG. Catheter ablation of the mitral isthmus for VT associated with inferior infarction. Circulation 1995;92.

10. de Bakker JM, van Capelle FJ, Janse MJ, Tasseron S, Vermeulen JT, de Jonge N, and Lahpor JR. Slow conduction in the infarcted human heart. 'Zigzag' course of activation. Circulation 1993;88:915-26.

11. Andresen D, Steinbeck G, Bruggemann T, Muller D, Haberl R, Behrens S, Hoffmann E, Wegscheider K, Dissmann R, and Ehlers HC. Risk stratification following myocardial infarction in the thrombolytic era: a two-step trategy using noninvasive and invasive methods. J Am Coll Cardiol 1999;33:131-8.

12. Gepstein L, Hayam G, and Ben-Haim S. A novel method for nonfluoroscopic catheter-based electroanatomical mapping of the heart: In vitro and in vivo accuracy results. Circulation 1997;95:1611-22.

13. Gardner MJ, Montague TJ, Armstrong CS, Horácek BM, and Smith ER. Vulnerability to ventricular arrhythmia: assessment by mapping of body surface potentials. Circulation 1986;73:684-92.

14. Horácek BM, Montague TJ, Gardner MJ, and Smith ER. Arrhythmogenic conditions. In DM Mirvis, editor, Body surface electrocardiographic mapping, pages 167-89. Kluwer Academic Publishers, Boston, MA, 1988.

15. Mitchell LB, Hubley-Kozey CL, Smith ER, Wyse DG, Duff HJ, Gillis AM, and Horácek BM. Electrocardiographic body surface mapping in patients with VT: Assessment of utility in the identification of effective pharmacological therapy. Circulation 1992;86:383-93.

16. Hubley-Kozey CL, Mitchell LB, Gardner MJ, Warren JW, Penney CJ, Smith ER, and Horácek BM. Spatial features in body-surface potential maps can identify patients with a history of sustained VT. Circulation 1995;92:1825-38.

17. SippensGroenewegen A, Spekhorst H, van Hemel NM, Kingma JH, Hauer RNW, Janse MJ, and Dunning AJ. Body surface mapping of ectopic left and right ventricular activation: QRS spectrum in patients without structural heart disease. Circulation 1990;82:879-96.

18. SippensGroenewegen A, Spekhorst H, van Hemel NM, Kingma JH, Hauer RNW, Janse MJ, and Dunning AJ. Body surface mapping of ectopic left ventricular activation: QRS spectrum in patients with prior myocardial infarction. Circ Res 1992;71:1361-78.

19. SippensGroenewegen A, Spekhorst H, van Hemel NM, Kingma JH, Hauer RNW, de Bakker JM, Grimbergen CA, Janse MJ, and Dunning AJ. Localization of the site of origin of postinfarction ventricular tachycardia by endocardial pace mapping: Body surface mapping compared with the 12-lead electrocardiogram. Circulation 1993 88:2290-306.

20. Hren R and BM Horácek . Value of simulated body surface potential maps as templates in localizing sites of ectopic activation for radiofrequency ablation. Physiol Meas 1997;18:373--400.

21. Lux RL, Evans AK, Burgess MJ, Wyatt RF, Abildskov JA. Redundancy reduction for improved display and analysis of body surface potential maps. I. Spatial compression. Circ Res 1981;49:186-96.

22. Evans AK, Lux RL, Burgess MJ, Wyatt RF, Abildskov JA. Redundancy reduction for improved display and analysis of body surface potential maps. II. Temporal compression. Circ Res 1981;49:197-203.

23. Gelbrands JJ. On the relationship between SVD, KLT and PCA. Pattern Recognition 1981;14:375-81.

24. SAS User’s Guide: Statistics. Cary, NC: SAS Institute Inc; 1982.

25. Efron B. The Jacknife, the Bootstrap, and Other Resampling Plans. Philadelphia, PA: Society for Industrial and Applied Mathematics; 1982:49-59