1. Introduction

Since the first detection of the magnetocardiogram (MCG) by Baule and McFee 1963 [Baule and McFee, 1963] biomagnetism, especially magneto encephalography (MEG), has been a subject of intensive research. Up to now the clinical results have not been very promising. In MCG the lack of success has been a consequence of inadequate recording methods.
Magnetic detection of the bioelectric activity includes both technical and bioelectromagnetic differences compared to the electric method.
One technical advantage of the magnetic method is that no electrodes are attached to the skin. Furthermore, superconducting SQUID detectors are capable of detecting DC-currents. On the other hand, biomagnetic instrumentation is much more expensive, especially if the magnetically shielded room is used.
The bioelectromagnetic differences are explained by the differences in the lead fields [Baule and McFee, 1970], i.e. measurement sensitivity distributions of electric and magnetic methods [Malmivuo and Plonsey, 1995; Malmivuo et al., 1997].

2. Theory of Biomagnetic Measurements

The signal detected by an electric lead is [Malmivuo and Plonsey, 1995]:

(1)

where: Φ_LE = electric scalar potential due to reciprocal energization of the electric lead. The quantity •Jⁱdv is the strength of the impressed current source and is called flow (or flux) source Ie. The signal detected by a magnetic lead is [Malmivuo and Plonsey, 1995]:

(2)

where: Φ_LM = magnetic scalar potential due to reciprocal energization of the magnetic lead. The quantity ×Jⁱ is defined as the vortex source Iv.

The geometric form of the sensitivity distributions of the recordings of the flow and vortex sources, i.e. the electric and magnetic lead fields, are shown in Fig. 1. They are independent on the basis of Helmholtz Theorem [Malmivuo and Plonsey, 1995]. However, the six signals, measured by the dipolar electric and magnetic leads cannot be completely independent as they all arise due to different aspects of the underlying current source. The dipolar electric measurement, VECG, shall detect about the same amount of information from the electric source as the dipolar magnetic measurement, VMCG. When combining these as VEMCG, the amount of detected information of the combined system should increase. This issue is the clinical application of magnetocardiography.

Figure 1. Dipolar electric and magnetic lead fields. (a) The equivalent electric dipole of a volume source is detected with a lead system which has three orthogonal homogeneous, linear lead fields. (b) The equivalent magnetic dipole is detected with a lead system which has three orthogonal lead fields, each being oriented tangentially to the symmetry axis and having intensity proportional to the radial distance from the symmetry (coordinate) axis.

3. Material

The test material of the clinical study consisted of 313 subjects including 154 normal, healthy subjects (85 male and 67 female, age 54±11 years), 90 patients with old inferior myocardial infarction (IMI), (73 male and 17 female, age 59±10 years) and 71 patients with old anteroseptal myocardial infarction (AMI), (59 male and 12 female, age 59±5 years).
The clinical diagnosis of myocardial infarction was based on a history of chest pain, a significant release of creatine kinase from the heart (CK>300 U/l and CK-B>10 %) and characteristic changes in the ECG in the acute phase of the infarction during the hospital period. The span between the infarction and the time of MCG and ECG measurements varied from one week to several years.
In 106 patients, localization of the infarction was based on one or more positive findings in the following ECG-independent diagnostic tests: (a) a rest perfusion defect in clinical cardiac isotope tomography (Tl-201 SPECT), (b) a hypo- or akinetic wall of the heart detected in echocardiography, or (c) an epicardial scar in the myocardium found during a coronary by-pass operation. In the remaining 55 patients, localization of the infarction was solely based on the diagnostic clinical 12-lead ECG changes measured in the acute phase of the infarction during the initial period of hospitalization.
The normal reference material consisted of apparently healthy subjects without a history of medical treatment or chest pain or other possibly heart-related symptoms, normal arterial blood pressure (systolic pressure <150 mmHg and diastolic pressure <90 mmHg) and normal heart sounds. The 12-lead ECG and the Frank VCG were recorded in all normal subjects. A duration of the QRS complex less than 110 ms was required to exclude subjects with incidental ventricular conduction defects from the normal material. Subjects with a mean QRS axis of -30o or less in the frontal plane were also excluded.

4. Methods

The equivalent electric and magnetic dipoles of the heart were measured with the Frank vector ECG (VECG) [Frank, 1956] and the asymmetric unipositional VMCG system, illustrated in Fig. 2a [Malmivuo, 1976; Barry et al., 1977; Malmivuo and Plonsey, 1995; Malmivuo et al., 1997] respectively. The methods and devices used have been described earlier [Nousiainen et al., 1994 a, b]. To increase the signal-to-noise ratio of the MCG signal, it was averaged over several (typically 20-40) cardiac cycles. Although the noise level in the ECG recording was low enough, the ECG was also time-averaged to assure a similar effect of data processing on both signals. Because the ECG and MCG were not recorded simultaneously, ECG lead II was used in both recordings as a common time reference.

The fundamental innovation applied in this study is the combination of the ECG and MCG recordings, called electromagnetocardiography (EMCG) [Malmivuo et al., 1997]. In this method diagnostic parameters from both the ECG and MCG are included in the analysis according to their diagnostic power. Because all the six dipolar electric and magnetic lead fields are mutually independent the available diagnostic information is maximized. Adding dipolar leads which are not independent of the existing ones but their linear combinations, does not increase diagnostic information [Willems et al., 1987].

Figure 2. Unipositional MCG lead system. (a) In the asymmetric unipositional lead system the MCG is recorded only on the anterior side of the thorax. (b) In the symmetric unipositional lead system the measurements are made on both sides of the thorax.

From the time-aligned averaged signals similar sets of parameters for ECG and MCG were extracted [Oja, 1993]. To compensate for the effect of interindividual variability in the QRS duration on the instantaneous wave amplitudes, the QRS and ST-T waves were time-normalized in each individual and the amplitude values were measured at 5% and 10% intervals throughout the duration of the QRS and T waves, respectively. The variable vector fed to the stepwise linear discriminant analysis (LDA) included 19 time normalized QRS and 9 ST?T amplitude values, maximum and minimum QRS?amplitudes and their time instants. Thus 32 parameters were calculated from each of the six dipolar leads being altogether 192 parameters. Only such parameters were used which can be derived from individual leads.

The statistical analysis of the data was performed with BMDP Statistical Software package [BMPD Manual, 1990]. The correct classification rate was evaluated with LDA by finding the linear combination of parameters (a classification function) that best predicts the group to which each individual case belongs [Mardia et al., 1989]. At the most eight best parameters of the dipolar ECG and MCG leads were used in the LDA. The classification results were assessed with the jackknife modification [Lachenbruch and Mickey, 1968; Oja et al., 1993]. This reduces the bias introduced by using the same data in learning and evaluating the classification which usually leads to overly optimistic results. The jackknife method was performed by leaving each case in turn out of the computation of the classification function and then using the function to classify the case omitted. The McNemar test of symmetry was used to test the significance of the observed differences in the classification [Bailey et al., 1998].

5. Results

We first tested the correct classification rates between normals and IMI patients. The correct classification rates for ECG, MCG and EMCG were 90.1%, 91.7 % and 95.5 %, respectively.
Table I compares the best classification rates of ECG and EMCG. From the 24 normal and IMI cases classified incorrectly by ECG, EMCG could classify correctly 19 cases. The best classification power increased by 5.4 % (from 90.1 % to 95.5 %, p = 0.015). Table II lists the parameters progressively added by the statistical software.
We made the same test for normals and patients with old anteroseptal myocardial infarction AMI. The correct classification rates for ECG, MCG and EMCG were 88.4 %, 87.4 % and 91.3 %, respectively.

TABLE 1. Frequency Table of classification results (number and %)
of N/IMI between ECG and EMCG.

		correct	false	total
EMCG	correct	212 (87.6%)	19 (7.9%)	231 (95.5%)
	false	6 (2.5%)	5 (2.1%)	11 (4.5%)
	total	218 (90.1%)	24 (9.9%)	242 (100%)

TABLE 2. Correct jackknifed classification rates between N/IMI patients with increasing number of parameters in the ECG, MCG and EMCG. The statistical significance, p, in the difference between the correct classification rates in the ECG and MCG is also given. (The parameters are progressively added as we go from top to bottom)

6. Conclusion

1. Clinically, the increased information in the EMCG method compared to ECG or MCG is manifested as a decrease of misclassified cases, in this study in N/IMI by about 50 %. This achieved improvement in the correct classification rate is statistically significant. This improvement is based on the fact that the patient sets classified correctly by ECG and MCG separately are approximately equal in size but not identical and the EMCG method can add the correctly classified cases from both sets.
   
2. Because different cardiac diseases are reflected in a different way in different leads, different leads are the best ones in diagnosing different cardiac diseases.
   
3. The issue, which method in general has better diagnostic power, the ECG or the MCG, is equally irrelevant as the issue, which lead, X, Y, or Z of the Frank VECG system is the best. The three plus three dipolar electric and magnetic leads belong to the same family of dipolar leads. In diagnosing one disease one of them is best, in another disease another one is best.
   
4. Which lead systems are used for detecting the VECG and VMCG is a practical question and of secondary importance as a theoretical aspect. In clinical practice the lead systems will be selected on the basis of clinical applicability, signal to noise ratio, etc.
   
5. Because MCG can be easily recorded in an unshielded environment, the price of the required instrumentation is not too high compared to the increased diagnostic power provided and therefore the MCG used in this context has clinical value.

7. Discussion

MCG Includes Information Partially Independent of ECG
We anticipate that an improvement in correct classification rate, similar to that obtained in this study at N/IMI and N/AMI, can also be achieved in most categories of cardiac diseases. Because some cardiac diseases, like arrhythmias, can be diagnosed even with one ECG with about 100% accuracy, the improvement given by MCG will be marginal.

Importance of Vectorial Measurements
Table II demonstrates clearly how important it is to make the dipolar MCG measurements in vectorial form, i.e. to measure all the three orthogonal dipolar components. In all the other MCG-studies in the world, only the component normal to the chest, the x-component, is recorded. We claim that this is one reason for the limited success of the other MCG studies.
In the clinical ECG, the measurement of all the three orthogonal dipolars is a routine procedure. This was introduced already in the 12-lead system by Wilson et al. in 1944 [Wilson et al., 1944]. The first accurate and clinically practical VECG system was introduced by Frank in 1956 [Frank, 1956]. The first proposal for an accurate and clinically practical VMCG system and the first recording of the vector magnetocardiogram was made by Malmivuo in 1976 [Malmivuo, 1976].

Redundancy of Additional Dipolar Leads
The amount of new information provided by the EMCG can be assessed against the information provided by the combination of the 12-lead and Frank VCG methods. Willems et al. [Willems et al., 1987] tested the performance of this combination to discriminate seven patient groups. They found that it did not improve the correct classification rate. The reason for this is that of these 15 leads only the three orthogonal ones are independent and the other 12 leads are redundant.

This fact is further proven by the well known works where the 12-lead ECG and VECG signals have been synthesized from each other with great accuracy. [Cady et al., 1966; Cady et al., 1971; Macfarlane and Lawrie, 1989]. If one set of signals is accurately synthesized from another set of signals, the information content of these signal sets is equal.

On this basis adding more dipolar magnetic leads to the three orthogonal ones would not increase the correct classification rate of MCG.

Further Improvements of the MCG Technique
The lead field of the asymmetric unipositional MCG lead system is not ideal for detecting the magnetic dipole moment of the heart, having its highest sensitivity to the electric sources located in the anterior region of the heart [Eskola et al., 1987; Malmivuo and Plonsey, 1995; Malmivuo et al., 1997]. We believe that recording the MCG on both sides of the thorax with the symmetric unipositional lead system [Malmivuo, 1976; Malmivuo and Plonsey, 1995; Malmivuo et al., 1997] would result in better correct classification rate of the MCG.

Other MCG Studies
There exist only a few comparative studies about the clinical performance of ECG and MCG. However, no study reports about combined analysis of the ECG and MCG. For example, Fujino et al. [Fujino et al., 1984] showed with 60 normals and 95 patients with left ventricular hypertrophy, LVH, that MCG detected with a 6x6 grid system could improve the diagnosis of LVH with respect to the 12-lead ECG from 76.1% to 79.6%.

In the sense of correct classification rate our results are superior to those of other studies. Also the construction of the instrumentation and the measurement protocol of the unipositional measurement system [Malmivuo and Plonsey, 1995; Malmivuo et al., 1997] are far simpler than those of the other studies. This makes the cost/efficiency ratio of our method superior to other proposed methods and we believe that this will lead to a widespread clinical application of the MCG.

Acknowledgements

This work has been supported by the Ragnar Granit Foundation.

References