JOURNAL OF ELECTROCARDIOLOGY, 21 (4), 1988, 361-367

Vectorcardiogram Synthesized From a la-lead ECG:

Superiority of the Inverse Dower Matrix
by LARS EDENBRANDT,M.D., AND OLLEPAHLM,M.D., PH.D

SUMMARY
Vectorcardiographic (VCG) criteria for the diagnosis of, for example, myocar-
dial infarction and right ventricular hypertrophy, are superior to the corre-
sponding la-lead ECG criteria. Contour and rotation of the QRS loops are im-
portant parts of these VCG criteria that have no direct counterpart in the la-lead
ECG. Therefore, attempts have been made to synthesize VCGs from la-lead ECGs
for diagnostic purposes. Visual comparison of QRS loops from the Frank VCG
and three different synthesized VCGs was made by three independent observers
to determine which method produces the most Frank-like QRS loops. The inverse
transformation matrix of Dower proved to be the best method of synthesis. Nor-
mal limits for some clinically important measurements in VCG interpretation
were calculated for this synthesis method and the Frank VCG.

Vectorcardiography (VCG) is superior to elec- plemented in the ECG management systems of

trocardiography (ECG) for the diagnosis of, for ex-
ample, myocardial infarction and right ventric-
ular hypertrophy.1-5 VCG criteria account for
configuration of the loops as well as measure-
ments of amplitude and duration, whereas con-
tour and rotation of the loops have no direct coun-
terpart in la-lead ECG.
Although the diagnostic superiority of the VCG
has been established for many years, the 12-lead
ECG is still most widely used. Attempts have been
made to incorporate the concepts of VCG into ECG
interpretation.6,7 One approach is to synthesize a
VCG from the 124ead ECG,* which enables loop
configuration to be used in the interpretation of
the ECG as it is used in VCG interpretation.
In this study, we compared three methods of
synthesizing VCG, using the Frank VCG as a ref-
erence. The first method is based on the Dower
transformation matrix,g the second is that de-
scribed by Bjerle et al.,* and the third is that im-
From the Departmentof Clinical Physiology,Universityof
Lund, Lund, Sweden.
Supportedby the National AssociationAgainst Chest and
Heart Diseases and the Swedish Medical ResearchCouncil,
Grant 02872.
Reprint,requests to: Lars Edenbrandt,M.D., Department
of Clinical Physiology, University Hospital, S-221 85 Lund,
Sweden.
Marquette Electronics Inc. (Milwaukee, W1j.l’
Our goal was to determine which method of syn-
thesis yields QRS loops with the greatest simi-
larity to the QRS loops of the Frank VCG and then
to present normal limits for the best method of
synthesis.
MATERIALS AND METHODS
RecordingTechniqueand SynthesisMethods
On each subject, a 12-lead ECG and a VCG, per-
formed according to the method of Frank,” were
recorded in rapid succession, with the subject in supine
position. The VCGs were recorded with the chest elec-
trodes in the fourth intercostal space.” Our recording
technique for VCG has been described elsewhere.13 All
ECG leads were recorded and A/D converted simul-
taneously. The synthesized VCG leads were then cal-
culated from the ECG leads, on a computer. After a
change of the coefficients in the equations, the synthe-
sized leads could be recalculated using the same ECG
recording. All VCGs and synthesized VCGs were plot-
ted on a Hewlett-Packard 7550A Graphics Plotter.
Three different synthesized VCGs (SVCG) were stud-
ied. The first SVCG, SVCG A, results from the inver-
sion of the equations proposed by Dower et al.’ for cal-
culating la-lead ECGs from Frank VCGs (for details,
see appendix). The resulting relationships are:

361
362 EDENBRANDT AND PAHLM
Xa = -0.172.V1 - 0.074.Vo + 0.122.Vo + 0.231.V4 + 0.239.V5 + 0.194.Vo + 0.156.1 - 0.010.11
Ya = 0.057.V1 - 0.019.Vz - 0.106.Vo - 0.022.V4 + 0.041.Vs + 0.043Vo - 0.227.1 + 0.887.11
ZA = -0.229.Vi - 0.31O.Va - 0.246,Vo - 0.063.V4 + 0.055.V5 + O.lOf+Vo + 0.022.1 + 0.102.11
The second SVCG, SVCG B, is based on the
equations proposed by Bjerle et al.‘:
Xn = 1.06.Va
Yn = 1.88*VF = 1.25aVF
Zs = 0.043*Vs - 0.532.Vz
The third SVCG, SVCG C, is the one that is
implemented in the ECG management systems of
Marquette Electronics Inc. (Milwaukee, WI)lO:
xc = I
Ye = aVF
Zc = -0.80’1 + VJ2)
Study Group
The comparison between the three SVCGs and the
Frank VCG was made on recordings from 80 subjects.
Fifty of the subjects were patients referred to the Clin-
ical Physiology Department for vectorcardiography and
30 subjects were selected at random from the normal
material, described below. Interpretation of the 80
Frank VCGs, considering established VCG criteria,2,4s5,
14,15yielded the following diagnoses: anterior myocar-
dial infarction, 9; inferior myocardial infarction, 16;
posterior myocardial infarction, 5; right ventricular hy-
pertrophy, 11; and normal findings, 39.
Normal Material
The subjects included in the normal material were
selected at random from a defined urban population.
Altogether, 472 subjects were examined and inter-
viewed. Those with known or suspected heart disease,
lung disease, or any other pathologic condition that
may influence the ECG or VCG were excluded, as were
six subjects with technically deficient ECG or VCG. A
total of 351 subjects (158 men and 193 women) aged
21-60 years were finally accepted for the study. The
material has been described in detail elsewhere.17
Visual Comparison of Methods
Visual comparison of the QRS loops from the Frank
VCG and the three SVCGs was made by two physicians
experienced in VCG interpretation and one technician
familiar with Frank VCGs. For each of the 80 Frank
VCGs, the three SVCGs were presented to the observ-
ers in random order with no indication of which method
of synthesis had been used. The observers decided, in-
dependent of each other, whether or not one of the
SVCGs showed greater agreement with the Frank VCG
than the other two.
Numerical Comparison and Normal Limits
Differences in measurements of the QRS and T loops
between the Frank VCG and the SVCGs were studied
in the normal material. Normal limits for some clini-
cally important measurements in VCG interpretation
were calculated for the best synthesis method and the
Frank VCG. The normal limits are defined as the 2.5
and 97.5 percentiles and are given separately for men
and women.
RESULTS
Figures 1 and 2 are examples of characteristic
differences between the Frank VCG and the
SVCGs. In a woman in the normal material (Fig.
l), the greatest differences in the scalar leads are
the lower amplitudes in lead Xc and the higher
amplitudes in Zc, compared with Frank leads X
and Z, respectively. These differences become
even more marked in the QRS loops. Differences
in the scalar leads between Frank VCG, SVCG A,
and SVCG B are not so obvious. The QRS loops
of SVCG A are, however, more Frank-like than
those of SVCG B.
Figure 2 shows QRS loops in the horizontal
plane of Frank VCG and SVCG A, B, and C. In
Frank VCG most of the QRS loop area is in the
right posterior quadrant. This is one of the mea-
surements used in the criteria for the diagnosis
of right ventricular hypertrophy.15 The Frank
VCG and SVCG A have similar configurations,
and in SVCG C, also, most of the QRS loop area
is in the right posterior quadrant. The greatest
difference is seen between the Frank VCG and
SVCG B, which has only a small S wave in lead
TABLE I
Ability of Methods A, B, and C to Synthesize a
Frank-like VCG, in 80 Subjects
SVCG
No SVCG
Observer A B C Selected Total
1 47 4 5 24 80
2 62 4 5 9 80
3 55 8 4 13 80
Total 164 16 14 46 240
Selection by three observers of the synthesized VCG
that agreed best with the Frank VCG. No SVCG was
selected when two of the SVCGs were equivalent and
agreed better with the Frank VCG than the third SVCG
or when all three SVCGs were equivalent.
 SYNTHESIZED VCG 363

FRANK SVCG A SVCG B SVCG C

XL XA--h- xBL XC-+-

YL- y*,h^_ ,L &-

Z--L ZA-L ‘B-k- zC -L

ImV
I, 0.5
sI
FRANK SVCG A SVCG B SVCG C

Fig. 1. Frank VCG and synthesizedVCGs A, B, and C in a normal subject. Scalar leads and QRS loops of the
horizontal, frontal, and left sagittal plane. Each loop has an indication at 25 msec and 50 msec after QRS onset.
Marks on the axes indicate 1 mV.

Xn and consequently does not produce a Frank-
like QRS loop in the horizontal plane.
Comparison of Methods for Synthesis
The results of visual comparison of the SVCGs
and the Frank VCG are shown in Table I. The
three observers selected SVCG A in about two
thirds of the cases. SVCG B and C were each se-
lected in less than one tenth of the cases.
The differences in measurements between the
SVCGs and the Frank VCG in the normal ma-
terial are shown in Table II. The differences in R
wave amplitudes in the scalar leads were smallest
for SVCG B and greatest for SVCG C. The R wave
amplitudes were 40% smaller in lead Xc and 50%
larger in lead Zc than in Frank leads X and Z,
respectively. Therefore, the maximal QRS vector
in the horizontal plane was on average 35” more
posteriorly directed in SVCG C than in Frank
VCG. In SVCG A and SVCG B, this vector was
on average only slightly more anteriorly directed
than in the Frank VCG. However, the maximal
T vector in the same plane was on average slightly
364 EDENBRANDTANDPAHLM

FRANK SVCG A SVCG B SVCG C

x- xA-+---- XB- xC----+--

zh ZA--h- ZBJ- z,-_,

1mV
I,0.5 s
i

Fig. 2. Horizontal QRS loops of Frank VCG and synthesizedVCGs A, B, and C in a patient with a Frank VCG
indicating right ventricular hypertrophy.Each loop has an indication at 25 msec and 50 msec after QRS onset.
Marks on the axes indicate 1 &V. - _

more anteriorly directed for all synthesis meth-
ods. Even though there are great differences be-
tween the SVCGs with regard to scalar ampli-
tudes, the maximal spatial amplitudes were about
0.1 mV larger in all SVCGs than in the Frank
VCG.
Normal Limits
The normal limits for measurements used in
the criteria for diagnosis of myocardial
infarction4s5*14and ventricular hypertrophy2,15,16
are presented in Table III. Most of the differences
between the limits in the Frank VCG and SVCG
A are small, but some may be of diagnostic im-
portance.
In women, more anteriorly directed vectors are
seen in SVCG A than in Frank VCG; this results
in a more anteriorly directed half-area vector in
the horizontal plane and a higher upper limit for
anterior duration. The upper limits of the maxi-
mal QRS vector in the horizontal plane are
greater in SVCG A for both men and women. Late
vectors directed to the right are less pronounced
in SVCG A than in the Frank VCG (Fig. 1). This
also influences the upper limits of the QRS loop
areas in the right posterior and right inferior
quadrants in the horizontal and frontal plane,
respectively.
DISCUSSION
One approach to incorporating the concepts of
VCG into the analysis of 12-lead EGC is to es-
tablish ECG criteria with new limits for ampli-
tudes and durations.7 The ECG is then inter-
preted by measuring amplitudes and durations for
1 of the 12 ECG leads at a time and comparing
them with thresholds. Another approach is to pre-
dict the configuration of the VCG loops from the
12-lead ECG.’ The contour and rotation of the
QRS loops are important elements of VCG criteria
that have no direct counterparts in 12-lead ECG.
Warner et a1.6studied temporal relationships be-
tween portions of the QRS recorded in lead II and
those recorded in lead III to predict the contour
of the frontal plane loop of the VCG. They pro-
posed new ECG criteria for inferior myocardial
infarction with performance similar to that of the
corresponding VCG criteria. These new ECG cri-
teria were more sensitive than conventional ECG
criteria.
The aim of this study was to find a method for
synthesizing a complete Frank-like VCG for di-
agnostic purposes. Therefore, the most important
part of the study was the visual comparison of the
SVCGs with the Frank VCG as a reference. In the
majority of cases, SVCG A proved to synthesize
the most Frank-like VCG. For some measure-
ments, such as R wave amplitudes in the scalar
leads, SVCG B showed greater similarities with
the Frank VCG than did SVCG A, but these dif-
ferences were usually small.
The fact that method A produces a more Frank-
like VCG than the other two methods does not
necessarily imply that the diagnostic information
content is greater in SVCG A than in the SVCG
B and SVCG C. However, one can apply experi-
 SYNTHESIZED VCG 365

TABLE II
Amplitudes and Directions in 351 Normal Subjects

Differences

Frank SVCG A-Frank SVCG B-Frank SVCG C-Frank

R, amplitude (mV) 1.02 f 0.34 0.20 k 0.22 0.12 r 0.27 -0.41 ? 0.19
R, amplitude (mV) 0.98 +- 0.35 -0.15 k 0.14 0.04 % 0.20 -0.17 t 0.14
R, amplitude (mV) 0.84 f 0.29 0.13 * 0.19 0.01 k 0.19 0.54 k 0.31
Max spatial QRS vector amplitude (mV) 1.57 2 0.38 0.12 k 0.24 0.14 t 0.29 0.09 * 0.36
Max horizontal QRS vector direction (“) 37 k 32 -6 t 25 -6 c 25 35 t 30
Max horizontal T vector direction (“) -31 k 18 -11 2 10 -14 * 11 -16 k 15

Values given are mean 4 SD.

ence from Frank VCGs in interpreting SVCGs if
the differences between them are small.
Figure 3 illustrates a situation in which SVCG
A can improve the diagnostic performance of
ECG. ECG, Frank VCG, and SVCG A are re-
corded for a patient with myocardial infarction
verified by myocardial scintigraphy. There are no
pathologic Q waves in the inferior ECG leads, but
the Frank VCG fulfills criteria for inferior myo-
cardial infarction.4 The configuration of the fron-
tal loop in SVCG A is very similar to that of the
Frank VCG. Interpreting SVCG A using Frank
VCG criteria would also yield the diagnosis of in-
ferior myocardial infarction.
The slightly more complex calculations used for
SVCG A than for SVCG B and SVCG C are not

TABLE III
Normal Limits for SVCG A and Frank VCG

SVCG A Frank

Men Women Men Women

Anterior Ml
Q, duration (msec) LL 22 21 25 22
Q, amplitude (mV) LL 0.1 0.1 0.1 0.1
Inferior Ml
Q, duration (msec) UL 33 28 30 26
Q, amplitude (mV) UL 0.2 0.2 0.2 0.2
C&,/R, amplitude ratio UL 0.2 0.2 0.2 0.2
Max frontal QRS vector direction (“) LL 5 11 13 23
Posterior Ml
Q, duration (msec) UL 52 52 51 42
Q, amplitude (mV) UL 0.9 0.7 0.8 0.6
Anterior accession time (msec) UL 38 37 38 32
Direction of half-area vector in the horizontal plane (“) LL -13 -14 -14 0
Left VH
Max horizontal QRS vector amplitude (mV) UL 2.6 2.1 2.1 1.7
Max horizontal T vector direction (“1 LL -78 -75 -67 -60
Right VH
Left posterior diagonal amplitude (LPD) (mV) LL 0.2 0.3 0.1 0.3
Q, + S, - LPD amplitude (mV) UL 0.4 0.2 0.6 0.1
Percent of the total QRS loop area in the:
Right inferior quadrant in the frontal plane UL 38 20 40 28
Right posterior quadrant in the horizontal plane UL 26 18 36 25
Right posterior and the two anterior quadrants in the UL 78 63 75 55
horizontal plane

LL, lower normal limit; UL, upper normal limit; MI, myocardial infarction; LVH, left ventricular hypertrophy;
RVH, right ventricular hypertrophy.
366 EDENBRANDTANDPAHLM

ECG FRANK SVCG A

I=-
50
25 4

aVF _

Ill --y- t

1mV 0.5 s
I, I

Fig. 3. Inferior leads II, aVF, and III and the frontal QRS loops of Frank VCG and SVCG A in a patient with
myocardial infarction, Each loop has an indication at 25 msec and 50 msec after QRS onset. Marks on the axes
indicate 1 mV.

noticeably more time-consuming with modern
computer technology. The SVCG A requires eight
simultaneous ECG leads, versus three for SVCG
B and four for SVCG C, but, ECG recorders that
collect all eight independent ECG leads simul-
taneously are now widely used.
Although there are great similarities between
SVCG A and the Frank VCG, the differences
make it necessary to study whether the limits in
the Frank VCG criteria must be adjusted for use
with SVCG A. We compared normal limits for
some clinically important measurements in VCG
interpretation and found that some adjustments
of VCG criteria probably are necessary. However,
before SVCG A criteria can be established and
resulting diagnostic performance presented,
SVCG A must be studied in well-defined patient
groups.
Acknowledgments: The authors are grateful to Dr.
B. Lundh, who provided the opportunity to study syn-
thesized VCG in normal material, and to A. Ek and H.
Grytzell, for excellent technical support.
REFERENCES
1. CHOUTC: When is the vectorcardiogram superior
to the scalar electrocardiogram? J Am Co11Cardiol
8:791, 1986
2. COWDERY CD, WAGNER GS, STARRJW ETAL:New vec-
torcardiographic criteria for diagnosing right ven-
tricular hypertrophy in mitral stenosis: comparison
with electrocardiographic criteria. Circulation
62:1026, 1980
3. HURDHP II, STARLING MR, CRAWFORD MH ET AL:
Comparative accuracy of electrocardiographic and
vectorcardiographic criteria for inferior myocardial
infarction. Circulation 63:1025, 1981
4. STARRJW, WAGNER GS, BEHARVS ET AL:Vectorcar-
diographic criteria for the diagnosis of inferior myo-
cardial infarction. Circulation 49:829, 1974
5. STARRJW, WAGNER GS, DRAFFIN RM ETAL:Vector-
cardiographic criteria for the diagnosis of anterior
myocardial infarction. Circulation 53~229, 1976
6. WARNERR, HILLNE, SHEEHE PR ET AL: Improved
electrocardiographic criteria for the diagnosis of in-
ferior myocardial infarction. Circulation 66:422,
1982
7. WARNERRA, REGERM, HILLNE ET AL:Electrocar-
diographic criteria for the diagnosis of anterior
myocardial infarction: importance of the duration
of precordial R waves, Am J Cardiol 52:690, 1983
8. BJERLEP, ARVEDSON 0: Comparison of Frank vec-
torcardiogram with two different vectorcardi-
ograms derived from conventional ECG-leads. Proc
Eng Found Conf 11:13,1986
9. DOWER GE, MACHADO HB, OSBORNE JA: On deriving
the electrocardiogram from vectorcardiographic
leads. Clin Cardiol 3:87, 1980
10. Marquette Electronics Inc: Personal communica-
tion.
11. FRANKE: An accurate, clinically practical system
for spatial vectorcardiography. Circulation 13:737,
1956
12. LANGNER PH, OKADARH, MOORE SR, FIESHL: Com-
parison of four orthogonal systems of vectorcar-
diography. Circulation 17:46, 1958
13. EDENBRANDT L, JONSON B, LUNDH B, PAHLM0: Sex-
and age-related normal limits for the QRS complex
in vectorcardiography. Clin Physiol 7:525, 1987
14. HOFFMAN I, TAYMORRC, MORRISMH, KI~~ELLI:
Quantitative criteria for the diagnosis of dorsal in-
 SYNTHESIZED VCG 367

farction using the Frank vectorcardiogram.Am 16. ROMHILT

DW, GREENFIELD
JC, ESTESEH: Vectorcar-
Heart J 70:295, 1965 diographic diagnosis of left ventricular hypertro-
15. CHOUTC, MASANGKAY MP, YOUNGR ET AL:Simple phy. Circulation 37:X, 1968
quantitative vectorcardiographic criteria for the di- 17. LUNDHB: On the normal scalar ECG: a new clas-
agnosis of right ventricular hypertrophy. Circula- sitication system considering age, sex and heart po-
tion 48:1262, 1973 sition. Acta Med Stand (suppl 691) 1984

APPENDIX

SVCG A is constructed with an inverse Dower Let x denote the vector of the three Frank VCG
transformation matrix. Dower et al.’ derived a 12- leads and v the vector of the eight independent
lead ECG from Frank leads X, Y, and Z. For ex- ECG leads. Then
ample, they found that:
Ax = v
V1 = -0.515.x + 0.157.Y - 0.917.2 Let
M = A’A
The 24 coefficients of leads X, Y, and Z that they
and
used to derive the eight independent leads (VI-
Vc, I and II) of a 12-lead ECG are given in the M-‘M = I
transformation matrix A: where I is the identity matrix.

- 0.515 0.157 -0.917 Then

0.044 0.164 - 1.387 x = Ix = M-lMx = M-lATAX = M-IA’+
0.882 0.098 - 1.277
1.213 0.127 - 0.601 where M -IAT is the inverse Dower transforma-
A= tion matrix presented in the text. Numerically, it
1.125 0.127 - 0.086
0.831 0.076 0.230 was found that:
0.632 - 0.235 0.059 -0.172-0.074 0.122 0.2310.2390.194 0.156 -0.010
M-'AT = 0.057 -0.019 -0.106 -0.022 0.0410.048 -0.227 0.667
0.235 1.066 - 0.132 -0.229 -0.310 -0.246 -0.063 0.0550.108 0.022 0.102