1. Introduction

The study of biologically generated magnetic fields basically started in 1963, when the magnetic field generated by the human heart was detected for the first time [1]. At the time, a large induction coil was used, while since the 1970ies superconducting quantum-interference devices (SQUIDs) have been the detectors of choice for biomagnetic fields [2]. Although such SQUID detectors are expensive to buy and operate, a quite active research community has sprung up around biomagnetism, with major areas of interest being the magnetic fields of the human heart (magnetocardiography, MCG) [5] and brain (magnetoencephalography, MEG) [3, 4].

Our main interest lies in the development of new techniques for sensitive optical magnetometry. We have concentrated on the magnetic field generated by the human heart because this is one of the strongest biomagnetic signals and therefore an obvious first target. Still, with a maximum amplitude of 100 pT the human MCG is extremely weak, about 10^-6 of the geomagnetic field and orders of magnitude weaker than typical stray fields in a normal environment. From the point of view of medical applications, a map of the cardiomagnetic field, taken during suitable intervals of the cardiac cycle, can help to detect pathological changes in the heart and its function. While biomagnetic diagnostics is still mostly an area of research and exploration, a recent review by cardiologists identifies various circumstances where the technique has great value [6]. In a related review [5], the same authors describe the instrumentation available for biomagnetic research to-date, stating that more convenient detectors will be needed before biomagnetic diagnostics can realize its full potential.

We believe that we can provide such a technique in the form of a newly developed laser-pumped cesium vapor magnetometer in a gradiometer configuration. As a demonstration of the technical feasibility and the future potential of our technique we have mapped the human cardiomagnetic field outside the chest with a time resolution of milliseconds and a spatial resolution of a few centimeters, typical of biomagnetic requirements.

2. Experimental method and setup

The magnetometer we developed for the present work is an optical/radio-frequency double resonance device operated in the phase-locked M_x-mode (Fig. 1) [7]. A resonant circularly polarized laser beam from an extended-cavity diode laser (λ=894 nm) optically pumps and thereby spin-polarizes a thermal vapor of cesium atoms, which is contained in a glass cell (2 cm long, 2 cm diameter) together with a buffer gas. The Larmor precession of the corresponding atomic magnetization around a bias field of B ₀=5µT, subtending an angle of 45° with respect to the laser beam direction, is resonantly driven by a radio-frequency (rf) field B ₁ at right angle to ${\overrightarrow{B}}_0$ . Due to this precession the optical transmission of the vapor changes periodically, in step (but not necessarily in phase) with the magnitude of the component of the magnetization along the laser beam direction. The modulation of the transmitted light power at the Larmor precession frequency is detected with a non-magnetic photodiode placed behind the vapor cell and demodulated with a lock-in amplifier synchronized to the rf. The phase of the response exhibits a steep linear slope at line center, so that a small change ΔB ₀ of the magnetic field results in a signal change that is directly proportional to the field change. When the output signal is fed back to control the driving rf frequency (phaselocked operation, dashed line in Fig. 1) the response bandwidth of the magnetometer is no longer limited by the magnetic-resonance linewidth, the same advantage as that of the self-oscillating configuration [7]. This is important because a response bandwidth of at least DC-40 Hz is required for a useful interpretation of cardiomagnetic signals.

 

Fig. 1. Schematic setup of the optically-pumped M_x-magnetometer in the phaselocked mode. The voltage-controlled oscillator (VCO) is locked to the phase signal (dashed line)

Download Full Size | PDF

When operated in this M_x-configuration the device acts as a scalar magnetometer whose signal is a function of |${\overrightarrow{B}}_0$|. As the MCG-induced field changes Δ$\overrightarrow{B}$ are much smaller than the offset field ${\overrightarrow{B}}_0$ applied in the z-direction, the signal changes of the magnetometer will be (to first order) proportional to ΔB_z only [8].

Since the biomagnetic fields are so weak, great care must be taken to reduce the influence of external stray fields, for instance line-frequency noise or that from electrical machinery in the vicinity. In this first demonstration setup, we have operated the device inside a partially shielded room, where line-frequency interference is reduced about 150-fold, to about 0.4 nT rms (root-mean-square) amplitude.

The required further reduction is achieved by an optical gradiometer [8]. Since the cardiomagnetic field decays rapidly with distance from the chest, while stray fields are rather homogeneous by comparison, a sensor placed near the chest detects both cardiomagnetic and ambient fields, while a second sensor 7 cm away responds mainly to ambient fields. The difference signal, which is free of the homogeneous component of the stray fields, is formed as follows. The distant magnetometer is operated in a phaselocked mode, by controlling the common radio frequency for both cells. This servo loop ensures that the instantaneous Larmor precession frequency of the atomic moments at the position of the distant sensor is always in resonance with the driving frequency; because of the common rf, homogeneous background fluctuations are corrected for in the sensor near the chest.

In this way, line-frequency interference in the MCG sensor can be reduced by about another two orders of magnitude, which is then sufficient to facilitate the direct visualization of the MCG on an oscilloscope. This additional noise reduction is limited by fluctuations of inhomogeneous stray fields, which are not compensated by the firstorder gradiometer. The overall compensation will be improved in a future design using higher-order gradiometers, as has been demonstrated for SQUID magnetometry [3].

3. Results

Detection noise at the input of the lock-in amplifier is typically about 5–10% above shot noise, giving an intrinsic sensitivity of $\frac{100\mathrm{fT}}{\sqrt{\mathrm{Hz}}}$ , which is in the range of high-T_c SQUIDs for biomedical applications.

The MCG data (Fig. 2(a)) obtained with an early version of this optical gradiometer allows one to distinguish the typical features, i.e., the QRS complex as well as the T wave (see inset in Fig. 2). This data is comparable in quality to earlier attempts at optical biomagnetometry [12, 13], work that comprised taking isolated MCG traces and that was soon discontinued (as far as we know). Our unfiltered and unaveraged data is already good enough to produce spatially resolved maps of the cardiomagnetic field for those instances of the cardiac cycle where the field is particularly strong, like the R peak [9] (for the conventional labelling of ECG and MCG features see Fig. 3). When for illustration purposes a sophisticated denoising algorithm based on state-space analysis [10] is applied to the data the curve represented in Fig. 2(b) is obtained [11], comparable in quality to similarly denoised SQUID data. This filtering procedure was not employed for any of the other results in this report.

 

Fig. 2. Time series of the magnetocardiogram of one of the authors. (a) Output data of the gradiometer. (b) The data filtered with a state-of-the-art denoising algorithm [10, 11]. This filtering procedure was not employed for any of the other results in this report.

Download Full Size | PDF

 

Fig. 3. All 63 MCGs (each averaged over 100–120 beats) taken on the grid above the chest, plotted together in order to show the range of signal amplitudes and shapes. The characteristic MCG features are by convention designated by the letters P, Q, R, S, and T.

Download Full Size | PDF

In order to obtain maps also for other instances during the heart cycle it is necessary to improve the signal quality, for example by averaging over several cycles. Since even at rest the heart beat is not perfectly regular one needs a way to trigger the signal. A simple 3-lead ECG-device was used for this purpose. Non-magnetic electrodes were connected to the left and right wrist of the subject, with a reference electrode connected to the left ankle.

Then 63 pairs of time series (100 s long) of the gradiometer signal and the ECG trace were recorded on a 8×8 grid of points over the chest separated by 4 cm each (one corner point could not be accessed due to geometrical constraints in our current setup). For each of those pairs, the positions of the local maxima of the second derivative of the ECG trace (i.e., the centers of the R peaks) were used in an off-line data treatment to identify the exact position of each beat. The MCG traces were then cut into (potentially overlapping) segments of 1000 data points (corresponding to 1 s of data) such that the R peak was at index 333 in each of the segments. Finally, the data points with the same index were averaged. It should be noted that the same triggering and averaging procedure is also necessary for SQUID-generated data. In Fig. 3, all 63 averaged MCGs are plotted on top of each other, thus showing the overall range and shape of the MCG signals.

 

Fig. 4. (QRS dynamics, Left and Center (1.5 MB); T dynamics, Right (2.4 MB)) Selected video frames taken from the movies accompaying this publication. The red area corresponds to maximum field pointing towards the chest (maximum ΔB_z), blue to the strongest field emerging from the chest (minimum ΔB_z). Within each movie, all frames have the same color scale and are oriented such that the top right corner corresponds to the grid point closest to the left shoulder. The black line connects the local maximum and minimum in each frame. LEFT: Field distribution at the Q dip. CENTER: Field distribution at the R peak. RIGHT: Field distribution at the crest of the T wave (different color scale).

Download Full Size | PDF

Data of this quality is good enough to document the dynamics of the cardiomagnetic field distribution in two dimensions. An intuitive way of presenting that information is to produce an iso-field map for each of the 1000 sample times and string them together into a movie of the spatio-temporal dynamics of the cardiomagnetic field. In order to get an impression of the quality of and the information provided by such a movie, we show three frames in Fig. 4, taken from the movies accompanying this publication. The movies itself cover, respectively, the region of the QRS complex, where the strongest signals are encountered, and of the T wave, a region of particular interest for cardiologists [14]. In the eye of a physicist, the field distributions look approximately like they were being produced by an effective current flowing along a finite-sized line segment oriented perpendicularly to the line connecting the field maximum and the field minimum in each frame. For the T-interval the angular orientation of this current has shown diagnostic value [14] superior to that from ECGs at rest for the detection of coronary arterial diseases. The angles obtained with our setup for our test subject (Fig. 5) are compatible with those obtained with SQUIDs for healthy persons [14].

4. Conclusion

Although intrinsically less sensitive than SQUID detectors, laser-pumped magnetometers can be used to record two-dimensional maps of human cardiomagnetic signals: the main problem is not sensitivity but noise suppression. The fact that OPMs are less expensive by more than one order of magnitude than SQUID detectors and that their operation is practically maintenance- and cost-free should contribute substantially to a broader acceptance of magnetocardiography as a diagnostic tool. We envision that, once cardiomagnetometry has found its place in medical routine, the standard sensors will not be SQUIDs but rather laser-pumped OPMs.

 

Fig. 5. Orientation of the line pointing from the local field minimum to the local field maximum, called the magnetic-field-map (MFM) orientation in [14]. Angles are shown only for those time intervals where both field minimum and maximum clearly stand out against the noise level. An angle of 0° corresponds to a line pointing towards the right of the frame, with angles increasing in the counter-clockwise direction.

Download Full Size | PDF