Circ J 2002; 66: 649–654

Electrocardiography Subtraction Method of Detecting Low- Amplitude, High-Frequency Components (L-HFCs) Within the QRS Complex of Patients With Ventricular

Toshihiko Nanke, MD; Kiyoshi Nakazawa, MD; Mariko Arai, MD; Shounosuke Ryu, MD; Tsuneharu Sakurai, MT; Naoki Matsumoto, MD*; Chuichi Sato, MD; Fumihiko Miyake, MD

A new method of extracting the low-amplitude and high-frequency components (L-HFCs) was developed and this study investigated its usefulness in 86 subjects: 28 normal subjects (group 1) and 58 patients with a previous (MI). The patients were classified into 3 groups: group 2 with 38 patients without ventric- ular tachycardia (VT), group 3 with 13 patients with non-sustained VT, and group 4 with 7 patients with sustained VT. The new (ECG) subtraction method, using a mathematical filtering procedure, was used instead of conventional complex filtering. The continuous L-HFCs within the QRS complex were analyzed using a Z-lead recording. The duration of the continuous L-HFCs was significantly longer in group 4 than in groups 1 (p<0.0001), 2 (p<0.0001) or 3 (p<0.05) with all 3 filters. The ECG subtraction method is a powerful and useful new technique for identifying patients at risk for either sustained or non-sustained VT after MI, and overcomes several of the problems with the conventional signal-averaging method. (Circ J 2002; 66: 649–654) Key Words: High-frequency components; Late potential; Myocardial infarction; QRS complex;

he high-frequency components (HFC) of the electro- dominate the energy of the QRS complex13 and mask the cardiography (ECG) have been classically described HFCs. Furthermore, the L-HFCs within the QRS complex T using a high-fidelity ECG, not by conventional 12- were undetectable in most of the previous studies that used lead ECG.1–3 Recent studies have described a method for the signal-averaging method. We report a novel ECG the extraction of the low-amplitude and HFC of the ECG subtraction method for extracting the L-HFCs from within (L-HFCs) for the terminal portion of the QRS complex or the QRS complex. the early portion of the ST segment of the signal-averaged ECG using high gain amplifications and signal-averaging techniques.4–8 These L-HFCs, so-called ‘late potentials’, Methods correspond to the delayed and fragmented endocardial or Subjects epicardial electrograms, and are believed to represent the The subjects consisted of 25 healthy volunteers (Healthy substrate for reentrant ventricular tachycardia (VT).4–12 The group) and 58 patients with a previous myocardial infarc- L-HFCs not only of the terminal portion of the QRS com- tion (MI): 21 consecutive patients with acute MI admitted plex, but also within the QRS complex have been detected to hospital between January 1995 and December 1997, and by analyzing the signals from catheter electrodes inserted 37 patients with VT or suspected VT who were admitted into ventricles.12 However, a potential problem with the between February 1990 and December 1994. The 58 patients signal-averaged ECG is the attenuation of the L-HFCs were classified into 3 groups from the results of routine because of trigger jitter of the biological signals7 and the ECG check-ups and clinical Holter ECG studies: 38 patients beat-to-beat variation in the L-HFCs caused by the complex without a history of VT (noVT-MI group), 13 patients with filtering and repeated summation-averaging required to clinically documented spontaneous non-sustained VT (non- improve the signal-to-noise ratio. Recent signal-averaged susVT-MI group) and 7 patients with clinically documented ECG studies have used a variety of high- and low-pass spontaneous sustained VT (susVT-MI group) (Table 1). filters, but as these conventional filters were designed only The Healthy group consisted of all males (mean age, 28±2 for the terminal portion of the QRS complex or for the years), the noVT-MI group had 33 males and 5 females early ST segment, some signal distortion and filter artifact (mean age, 60±12 years; 21 anterior MI and 17 inferior MI), have been unavoidable. The low-frequency components the non-susVT-MI group had all males (mean age, 52±13 years; 10 anterior MI and 3 inferior MI), and the susVT-MI (Received December 27, 2001; revised manuscript received March group comprised 6 males and 1 female (mean age, 61±8 18, 2002; accepted April 1, 2002) years; 3 anterior MI and 4 inferior MI). The left ventricular Division of Cardiology, Department of Internal Medicine and *Depart- ejection fraction was significantly less in the susVT-MI ment of Pharmacology, St Marianna University School of Medicine, group (mean 36.2±5.6%, range 30–48%) than in the noVT- Kawasaki, Japan Mailing address: Toshihiko Nanke, MD, Division of Cardiology, MI group (mean 53.3±12.6%, range 32–78%) and the non- Department of Internal Medicine, St Marianna University School of susVT-MI group (mean 48.0±8.7%, range 34–61%). There Medicine, 2-16-1 Sugao, Miyamae-ku, Kawasaki 216-8511, Japan. was no significant difference in left ventricular ejection E-mail: [email protected] fraction between the susVT-MI and non-susVT-MI groups.

Circulation Journal Vol.66, July 2002 650 NANKE T et al.

Table 1 Clinical Data

Healthy group noVT-MI group non-susVT-MI group susVT-MI group (n=25) (n=38) (n=13) (n=7) Age (years) 28±2 60±12 52±13 61±8 Male/female 25/0 33/5 13/0 6/1 Myocardial infarction –/– 21/17 10/3 3/4 (anterior/inferior) LVEF (%) – 53.3±12.6 48.0±8.7 36.2±5.6 VT, ventricular tachycardia; MI, myocardial infarction; LVEF, left ventricular ejection fraction.

Non-sustained VT was defined as 5 or more consecutive original signals (Fig1). The first step was to make the low- premature ventricular complexes at a rate of more than 120 frequency components [B] from the original waves [A] beats/min lasting less than 30s. VT lasting more than 30s using signal smoothing with a moving average method. For was classified as sustained. Cases of acute MI within 4 the second step, the low-frequency components [B] were weeks prior to the study, , accessory subtracted from the original waves [A], which allowed only pathways, or atrial fibrillation were excluded. None of the the L-HFCs to remain. For the last step, the L-HFCs were patients were on antiarrhythmic medication at the time of amplified and smoothed by a high-cut filter to reduce the the study. Most of the patients were observed by cardiolo- noise in the super high-frequency zone to obtain the sub- gists at St Marianna University Hospital or affiliated hospi- traction ECG [C]. Changing the filter characteristics, such tals every 2–4 weeks, with a mean follow up period of 45 as the number of smoothing times, subtraction times, point months (range: 2–110 months). Informed consent to partic- of the moving average and final smoothing times, were ipate in the study was obtained from all the subjects. very easy with this method. Finally a magnitude display of the subtraction ECG [D] was calculated for each point of ECG Subtraction Method the subtraction wave by the following formula: In the ECG subtraction method, a mathematical filtering Magnitude display of the subtraction ECG (t) = Z (t)2 procedure was used instead of the conventional complex filtering, to extract the low-frequency components from the Data Acquisition Recording of data was performed in an electromagnetic shield room. The hair on the chest was shaved as needed for proper electrode attachment. After thorough cleaning and mild abrasion of the skin with a skin cleanser, silver/ silver chloride ECG electrodes (Magnerode, Fukuda Denshi, Inc, Tokyo, Japan) were applied for a Z-lead terminal at the fifth intercostal space along the mid-clavicular line (posi- tive electrode) and at an opposite point on the posterior chest (negative electrode). The original ECG signals were obtained using a bipolar Z-lead, amplified approximately 4,000-fold by a preamplifier (LA-804, Fukuda Denshi, Inc) and digitized at 5,000samples/s by a digital oscilloscope (310B, Nicolet, Inc, WI, USA) with 12-bit accuracy. The digital information stored on a floppy disk was analyzed by the ECG subtraction method with a personal computer system (Vectra-D, Hewlett-Packard, Inc, Sunnyvale, USA and Waveform BASIC) and the calculated wave was printed by an XY plotter (7470A, Hewlett-Packard, Inc). In all of the patients, the data from 5 beats were averaged and processed. The frequency characteristics of the filters were as follows (Fig2): the H1-filter ranged from 100 to 450Hz, and the peak voltage response was obtained at 160 Hz; the H2-filter ranged from 120 to 450Hz, and the peak Fig1. Technical procedure for the ECG subtraction method. voltage response was obtained at 200 Hz; the H3-filter

Fig2. Frequency characteristics of the H1-filter (Left), H2-filter (Middle) and H3-filter (Right).

Circulation Journal Vol.66, July 2002 ECG Subtraction Method for L-HFCs 651

Fig 3. A subtraction ECG obtained from a normal subject in the Healthy group. The unfiltered waveforms are shown in the top panel, the subtracted ECG in the middle panel, the magnitude display of the subtracted ECG in the third panel, and the H1-filtered HFCs on the left, H2-filtered HFCs in the middle and H3-filtered HFCs on the right. The L-HFCs were not sustained in any of the 3 filters (range, 13.9–19.0ms).

Fig4. A subtraction ECG obtained from a patient with an inferior myocardial infarction and subsequent sustained ventricular tachycardia (VT) in the susVT-MI group. The unfiltered waveforms are shown in the top panel, the subtracted ECG in the middle panel, the magnitude display of the subtracted ECG in the third panel, and the H1-filtered HFCs on the left, H2-filtered HFCs in the middle and H3-filtered HFCs on the right. Well-defined continuous L-HFCs (arrow) were present in the recording obtained from the H1, H2 and H3- filters (range, 65.5–70.2ms).

Fig5. A subtraction ECG obtained from a patient with an inferior myocardial infarction and non-sustained ventricular tachycardia (VT) in the non-susVT-MI group. The unfiltered waveforms are shown in the top panel, the subtracted ECG in the middle panel, the magnitude display of the subtracted ECG in the third panel, and the H1-filtered HFCs on the left, H2-filtered HFCs in the middle and H3-filtered HFCs on the right. The H3-filter clearly demonstrates continuous L-HFCs (42.9ms) starting from the QRS complex; however, L- HFCs were not detectable (range, 9.7–32.1ms) with the H1 or H2-filters. ranged from 160 to 450Hz, and the peak voltage response the 80-times smoothed original signals. This subtraction was obtained at 250Hz. The following mathematical pro- process was repeated once, then finished with one smooth- cedures were applied to obtain those filter characteristics. ing. (1) For the H1-filtering, the signals were subtracted by (3) For the H3-filtering, the signals were subtracted by the 40-times smoothed original signals, then finished with the 20-times smoothed original signals. This subtraction one smoothing. process was repeated for 20-times, then finished with one (2) For the H2-filtering, the signals were subtracted by smoothing.

Circulation Journal Vol.66, July 2002 652 NANKE T et al.

Fig 6. The duration of continuous L-HFCs for the H1-filter was Fig7. The duration of continuous L-HFCs duration for the H2-filter significantly longer in the susVT-MI group than in groups Healthy, was significantly longer in susVT-MI group than in groups Healthy, noVT-MI or non-susVT-MI. There was no significant difference in noVT-MI or non-susVT-MI. The continuous L-HFCs duration was the continuous L-HFCs duration among groups Healthy, noVT-MI or significantly longer in non-susVT-MI group than in Healthy group. non-susVT-MI. VT, ventricular tachycardia. VT, ventricular tachycardia.

less than 0.05 was considered statistically significant. The duration of the L-HFCs between the clinical groups were compared with nonparametric Kruskal-Wallis testing, and posthoc testing was performed with Scheffe’s F test.

Results Subtraction ECG of Representative Cases Representative results of subtraction ECGs from a normal healthy subject (Fig3), a patient with sustained VT (Fig4) and a patient with non-sustained VT (Fig5) are presented. The L-HFCs were not sustained after passing any of the 3 filters for all of the normal healthy subjects (range, 10.7– 39.8ms). All 7 patients with sustained VT had well-defined continuous L-HFCs starting from inside the QRS complex (as shown in Fig4). The continuous L-HFCs were repro- ducibly present in all recordings obtained with the 3 filters (H1, H2, H3) in the patients with sustained VT (range, Fig 8. The duration of continuous L-HFCs for the H3-filter was 33.3–72.6ms with H1-filter, 46.5–72.6ms with H2-filter significantly longer in group susVT-MI group than in groups Healthy, noVT-MI or non-susVT-MI. The continuous L-HFCs duration was and 52.4–78.6 ms with H3-filter); however, the precise significantly longer in non-susVT-MI group than in Healthy group. amplitude and morphology of the L-HFCs varied slightly VT, ventricular tachycardia. from day to day even in the analyses from the same subject. In contrast, well-defined continuous L-HFCs from the signals obtained from all 3 filter recordings were rare (1/13: The signal was considered to be significant when it was 8%) in the patients with non-sustained VT. In a representa- more than twice the size of the noise level of the subtracted tive analysis from a patient with non-sustained VT (Fig5), ECG. The noise level was measured during the ST segment the H3-filter clearly demonstrated continuous L-HFCs start- over a 70ms sample from the magnitude display of the ing from the beginning of the QRS complex; however, they subtraction ECG. The L-HFCs were defined as a low-ampli- were undetectable with the H1- and H2-filters. Continuous tude signal (<10μV) recorded from the terminal portion of L-HFCs were detected in 6 of 13 patients with non-sustained the subtraction ECG. The mean noise level obtained for the VT with the H3-filter, but not with the other filters. H1-filter was 0.53±0.13μV, 0.53±0.14μV for the H2-filter and 0.49±0.13μV for the H3-filter in the normal subject Comparison of the Duration of the Continuous group. The duration of the L-HFCs and noise level were L-HFCs for Each Group measured with the H1-, H2- and H3-filters. The duration of the continuous L-HFCs with the H1-filter (Fig 6) was significantly longer in the susVT-MI group Statistical Analysis (54.1±13.8 ms) than in the Healthy group (26.5±7.2 ms, All statistical analyses were performed with Statview p<0.0001), non-VT-MI group (26.0±13.9ms, p<0.0001) or software (Abacus Concepts, Inc, Calabasus, CA, USA). All non-susVT MI group (33.3±16.9ms, p=0.01). data are presented as mean ±1SD. A calculated p-value The duration of the continuous L-HFCs with the H2-filter

Circulation Journal Vol.66, July 2002 ECG Subtraction Method for L-HFCs 653

(Fig 7) was significantly longer in the susVT-MI group averaged than the conventional signal-averaging method: (54.4±8.8 ms) than in the Healthy group (24.7±6.0 ms, in the present study an analysis of the recordings from as p<0.0001), noVT-MI group (29.1±12.0ms, p<0.0001) or few as 5 beats successfully revealed L-HFCs. Secondly, non-susVT-MI group (37.3±18.5 ms, p=0.006). Further- there is an extremely low level of noise. The amplitude of more, the duration of the continuous L-HFCs with the H2- the clinically important signals (ie, the continuous L-HFCs filter was significantly longer in the non-susVT-MI group in patients with VT) is only 10μV, and the accepted guide- than in the Healthy group (p=0.024). line for signal-averaging is 0.7μV.23 However, with our The duration of the continuous L-HFCs with the H3- method, the noise levels were only from 0.49–0.53μV for filter (Fig8) was significantly longer in the susVT-MI group all 3 filters, which allowed us to clearly distinguish the (61.4±9.8 ms) than in the Healthy group (26.5±7.3 ms, signals from noise. In order to reduce the noise, we recorded p<0.0001), noVT-MI group (34.2±12.4ms, p<0.0001) or the original signals from the Z-lead with careful application non-susVT-MI group (44.6±17.1 ms, p=0.034). Further- of the electrodes so that the L-HFCs are hardly attenuated more, the duration of the continuous L-HFCs with the H3- and respiratory movement and electromyograms have little filter was significantly longer in the non-susVT-MI group effect on the recordings.24 Thirdly, conventional complex than in the Healthy group (p=0.0004). There was no signif- filtering is not necessary, which allows us to avoid a possi- icant difference in the duration of the continuous L-HFCs ble source of signal distortion and filter artifact. Finally, the between the noVT-MI group and non-susVT-MI group for filter characteristics can be easily modified by simply the H1-filter (p=0.377), H2-filter (p=0.198) or H3-filter changing the number of smoothing passes and subtractions, (p=0.064). as well as the number of moving average points. We tried several patterns of filtering maneuvers to determine the optimal characteristics of the filters to detect the L-HFCs not Discussion only in the terminal portion of the QRS complex as reported The high-frequency (>80Hz) components of the QRS in the previous studies, but also within the QRS complex.25,26 complex are less than 3% of the total voltage,13 but some The advantages of the ECG subtraction method enabled us investigators have suggested that these components might to detect continuous L-HFCs starting from the inside the be helpful in the prediction of VT in certain patients.14–17 In QRS complex and so we believe our method is better than previous studies, the HFCs were estimated by counting the the conventional signal-averaging method. number of notches and slurs of the QRS complex on a With regard to clinical application, the continuous L- high-fidelity ECG,1–3 but recently, high gain amplifications HFCs were present in all recordings made with the 3 filters and signal-averaging techniques have been reported in the in the patients with sustained VT, whereas they were absent evaluation of the quantitative magnitude of the L-HFCs in the normal healthy subjects. The duration of the continu- of the terminal portion of the QRS complex or the early ous L-HFCs in the patients with subsequent sustained VT portion of the ST segment on the signal-averaged ECG.4–8 was significantly longer than in the non-sustained VT The L-HFCs of the terminal portion of the QRS complex or patients, patients without VT or the normal healthy subjects. the early portion of the ST segment are called ‘late poten- From these results, the patients with sustained VT can quite tials’ and are associated either directly or indirectly with easily be differentiated. The frequency band of the late the presence of a reentry substrate for VT;9,10,12 it has been potential was reported by Cain et al to have a peak below suggested that these signals could be a non-invasive predic- 120Hz;7 however, we have more distinctive features with tor of sustained VT11,12,18,19 or of the risk for sudden cardiac the sharpest low-cut H3-filter. Some of the present patients death,19–21 particularly in patients with a history of MI. with non-sustained VT had well-defined continuous L- The signal-averaged ECG is a well-established tech- HFCs only with the H3-filter and they were undetectable nique for improving the signal-to-noise ratio on the body with the H1- and/or H2-filters. Those results suggest that surface ECG. The level of noise will be reduced by a factor the details of the HFCs within the QRS complex were clar- of 1/ N , where N is the number of averaged beats. The ified by our new method and we could differentiate patients typical signal-averaged ECG is obtained from the analysis with and without VT by the detection of L-HFCs using the of 100–600 beats, and no less than 64 beats.22 The conven- H3-filter. The following 3 mechanisms may be involved: tional signal-averaged ECG studies have some inherent (1) low-frequency components may dominate the energy of limitations for the analysis of the L-HFCs. First, the HFCs the QRS complex, masking the continuous L-HFCs, (2) al- tend to be attenuated because of trigger jitter, which can teration of the HFCs is the remains of the QRS complex and operate as a low-pass filter and masks the L-HFCs. Even (3) the alteration of the signal bands of the HFCs is caused with a 100-beat signal-averaged ECG system with a refer- by variation in the individual’s myocardial substrate.27 ence jitter of ±1ms, the –3dB cut-off point falls at 134Hz.6 Our new ECG subtraction method appears to be a pow- Other investigators have reported a jitter value of 0.5–2.0 erful and useful new technique for identifying patients at ms.4,6,11,18 These observations suggest that high-frequency risk for sustained VT and non-sustained VT after MI, and signals greater than 140Hz are likely to be attenuated. overcomes several of the problems of the conventional Secondly, repeated signal averaging causes somewhat vari- signal-averaging method. Further prospective studies with able L-HFCs because of the beat-to-beat variation. The a large number of patients are warranted. L-HFCs may drift third limitation is the difficulty in changing the filter char- with time after MI, with circadian rhythm or with day-to- acteristics. Previous signal-averaged ECG studies utilized day changes in condition, so further and detailed analysis is various high- and low-pass filters, but these filters were needed. designed to detect the L-HFCs in the terminal portion of the QRS complex and were not suitable for detection of the References L-HFCs within the QRS complex. 1. Reid WD, Caldwell SH. Research in electrocardiography. Ann Intern In contrast, our method has 4 novel aspects. First, the Med 1933; 7: 369–380. subtraction ECG method requires fewer ECG signals to be 2. Langner PH. The value of high fidelity electrocardiography using the

Circulation Journal Vol.66, July 2002 654 NANKE T et al.

cathode ray oscillograph and an expanded time scale. Circulation 1981; 64: 34–42. 1952; 5: 249–256. 17. Abboud S, Belhassen B, Miller HI, Sadeh D, Laniado S. High 3. Langner PH. Further studies in high fidelity electrocardiography: frequency electrocardiography using an advanced method of signal Myocardial infarction. Circulation 1953; 8: 905–913. averaging for non-invasive detection of in 4. Fontaine G, Frank R, Gallais-Hamonno F, Allali I, Phan-Thuc H, patients with normal conventional electrocardiogram. J Electrocardiol Grosroreat Y. Electrocardiography of delayed potentials in post- 1986; 19: 371–380. excitation syndrome. Arch Mal Coeur Vaiss 1978; 71: 854–864. 18. Rozanski JJ, Mortara D, Myerburg RJ, Castellanos A. Body surface 5. Berbari EJ, Scherlag BJ, Hope RR, Lazzara R. Recording from the detection of delayed depolarizations in patients with recurrent body surface of arrhythmogenic ventricular activity during the S-T ventricular tachycardia and left ventricular aneurysm. Circulation segment. Am J Cardiol 1978; 41: 697–702. 1981; 63: 1172–1178. 6. Breithardt G, Becker R, Seipel L, Abendorth RR, Ostermeyer J. 19. Breithardt G, Borggrefe M. Recent advances in the identification of Non-invasive detection of late potentials in man: A new marker for patients at risk of ventricular tachyarrhythmias: Role of ventricular ventricular tachycardia. Eur Heart J 1981; 2: 1–11. late potentials. Circulation 1987; 75: 1091–1096. 7. Cain ME, Ambos HD, Witkowski FX, Sobel BE. Fast-Fourier trans- 20. Kuchar DL, Thorburn CW, Sammel NL. Late potentials detected after form analysis of signal-averaged electrocardiograms for identification myocardial infarction: Natural history and prognostic significance. of patients prone to sustained ventricular tachycardia. Circulation Circulation 1986; 74: 1280–1289. 1984; 69: 711–720. 21. Denniss AR, Richards DA, Cody DV, Russell PA, Young AA, 8. Haberl R, Jilge G, Pulter R, Steinbeck G. Comparison of frequency Cooper MJ, et al. Prognostic significance of ventricular tachycardia and time domain analysis of the signal-averaged electrocardiogram and induced at programmed stimulation and delayed in patients with ventricular tachycardia and coronary artery disease: potentials detected on the signal-averaged electrocardiograms of Methodologic validation and clinical relevance. J Am Coll Cardiol survivors of acute myocardial infarction. Circulation 1986; 74: 731– 1988; 12: 150–158. 745. 9. El-Sherif N, Scherlag BJ, Lazzara R, Hope RR. Re-entrant ventricular 22. Lander P, Berbari EJ, Lazzara R. Optimal filtering and quality control in the late myocardial infarction period. 1. Conduction of the signal-averaged ECG: High-fidelity 1-minute recordings. characteristics in the infarction zone. Circulation 1977; 55: 686–702. Circulation 1995; 91: 1495–1505. 10. El-Sherif N, Hope RR, Scherlag BJ, Lazzara R. Re-entrant ventricu- 23. Breithardt G, Cain ME, El-Sherif N, Flowers NC, Hombach V, Janse lar arrhythmias in the late myocardial infarction period. 2. Patterns of M, et al. Standards for analysis of ventricular late potentials using initiation and termination of re-entry. Circulation 1977; 55: 702–719. high-resolution or signal-averaged electrocardiography. J Am Coll 11. Simson MB. Use of signals in the terminal QRS complex to identify Cardiol 1991; 17: 999–1006. patients with ventricular tachycardia after myocardial infarction. 24. Nanke T, Hara M, Sakurai T, Nakamura T, Tochiki H, Nakazawa K, Circulation 1981; 64: 235–242. et al. Development of a method to detect the high frequency compo- 12. Simson MB, Untereker WJ, Spielman SR, Horowitz LN, Marcus nent of the QRS complex by using the subtraction method and its NH, Falcone RA, et al. Relation between late potentials on the body application for myocardial infarction. Jpn Heart J 1994; 35(Suppl): surface and directly recorded fragmented electrograms in patients 319–320. with ventricular tachycardia. Am J Cardiol 1983; 51: 105–112. 25. Nanke T. A study on detecting high frequency components of QRS 13. Scher AM, Young AC. Frequency analysis of the electrocardiogram. complex by subtraction method and its application for intraventricu- Circ Res 1960; 8: 344–346. lar conduction disturbance. Jpn J Electrocardiol 1993; 13: 301–311. 14. Flowers NC, Horan LG, Tolleson WJ, Thomas JR. Localization of 26. Nanke T, Tochiki H, Hara M, So T, Sakurai T, Nakamura T, et al. A the site of myocardial scarring in man by high-frequency components. study on high frequency component in the QRS complex in reference Circulation 1969; 40: 927–934. to myocardial infarction site. Jpn J Electrocardiol 1995; 15: 245– 15. Langner PH, Geselowitz DB, Briller SA. Wide band recording of the 252. electrocardiogram and coronary heart disease. Am Heart J 1973; 86: 27. Hara M, Tuchiya K, Nanke T, Mori N, Miyake F, Sato C, et al. 308–317. Development of a new analytical method for the electrocardiogram 16. Goldberger AL, Bhargava V, Froelicher V, Covell J. Effect of using short-time first fourier transforms. Jpn Circ J 1999; 63: 941– myocardial infarction on high-frequency QRS potentials. Circulation 944.

Circulation Journal Vol.66, July 2002