Elsevier

NeuroImage

Volume 21, Issue 3, March 2004, Pages 1009-1020
NeuroImage

Human fetal brain imaging by magnetoencephalography: verification of fetal brain signals by comparison with fetal brain models

https://doi.org/10.1016/j.neuroimage.2003.10.022Get rights and content

Abstract

Fetal magnetoencephalogram (fMEG) is measured in the presence of a large interference from maternal and fetal magnetocardiograms (mMCG and fMCG). This cardiac interference can be successfully removed by orthogonal projection of the corresponding spatial vectors. However, orthogonal projection redistributes the fMEG signal among channels. Such redistribution can be readily accounted for in the forward solution, and the signal topography can also be corrected. To assure that the correction has been done properly, and also to verify that the measured signal originates from within the fetal head, we have modeled the observed fMEG by two extreme models where the fetal head is assumed to be either electrically transparent or isolated from the abdominal tissue. Based on the measured spontaneous, sharp wave, and flash-evoked fMEG signals, we have concluded that the model of the electrically isolated fetal head is more appropriate for fMEG analysis. We show with the help of this model that the redistribution due to projection was properly corrected, and also, that the measured fMEG is consistent with the known position of the fetal head. The modeling provides additional confidence that the measured signals indeed originate from within the fetal head.

Introduction

Investigations of fetal developmental brain processes are limited by the inaccessibility of the fetus. However, two techniques for the study of fetal brain function in utero are emerging: functional magnetic resonance imaging (fMRI) Hykin et al., 1999, Moore et al., 2001 and fetal magnetoencephalography (fMEG) Eswaran et al., 2002a, Eswaran et al., 2002b, Lengle et al., 2001, Schneider et al., 2001. There are several advantages and disadvantages of both techniques. fMRI has inherent limitations based on difficult access to the measuring space, high sound levels, and safety issues, but delivers both the functional and the anatomical information. In contrast, fMEG is a completely passive and noninvasive method with superior temporal resolution, but does not directly provide any anatomical information. This additional information has to be obtained by complementary imaging techniques, like ultrasound. In this paper, we investigate how the fMEG signals are validated by modeling and correlation with known positions of the fetal head.

The fetal magnetoencephalogram (fMEG) is measured in the presence of environmental noise and various near-field biological signals and other interferences, for example, maternal magnetocardiogram (mMCG), fetal magnetocardiogram (fMCG), uterine smooth muscle (magnetomyogram), and motion artifacts Robinson et al., 2002, Wakai and Lutter, 2002. After cancellation of environmental noise (Vrba, 2000), the mMCG and fMCG are usually the dominant artifacts and must be removed to observe fMEG. The magnitude of the averaged evoked fMEG signals is typically in the range from 10 to 80 fT Eswaran et al., 2002a, Lengle et al., 2001, while the fMCG and mMCG at the fetal thorax location can both attain amplitudes as large as 10 pT. Closer to the maternal heart, the mMCG can be as large as 100 pT.

The auditory evoked fMEG was first demonstrated by Blum et al. (1985) using a single-channel superconducting quantum interference device (SQUID) system. The analysis used the standard stimulus-triggered averaging and did not take into account the possible mMCG and fMCG interference. Thus the averaged evoked response was very noisy and difficult to interpret. Reduction of the MCG interference has become more practical with the advent of multichannel SQUID systems. In the majority of reported work, the MCG was reduced by adaptive filtering or noise estimation techniques Samonas et al., 1997, Strohbach et al., 1994. Samonas et al. (1997) compared different algorithms for elimination of MCG from MEG recordings: direct subtraction (DS) of a MCG signal, adaptive interference cancellation (AIC), and orthogonal signal projection algorithms (OSPA). The OSPA was found to be superior over other approaches as it minimally affected the brain signals while being robust to nonstationary MCG interference. All these approaches and their slightly modified versions are used for fMEG detection. The DS was applied by Schneider et al. (2001) for mMCG and fMCG cancellation. Zappasodi et al. (2001) used both AIC and OSPA with no clear cut difference in the performance of both cancellation algorithms. Typically, the MCG cancellation methods utilize detector arrays which permit both the spatial and temporal data manipulation of signal and interference Chen et al., 2001, Lengle et al., 2001, Wakai and Lutter, 2002 and utilize matched filtering or matched filtering assisted by spatial filtering. These methods separate the signal space into signal and noise subspaces (where the noise subspace includes the unwanted interference) and construct filters, which preserve signal subspace, while the contribution of the noise subspace to the final signal is greatly reduced. To date, the temporal morphology of the recorded signals was usually used as evidence that they originate within the fetal brain. Also, signals measured only by sensors located close to the fetal head were reported. It is shown in this paper that both spatial and temporal information are important for the identification of fMEG signals and that if orthogonal projection is used for mMCG and fMCG cancellation, signals at larger distances from the fetal head should also be measured.

We remove the fMCG and mMCG interference by projecting them out of the data by orthogonal projection Huotilainen et al., 1995, Tesche et al., 1995, Uusitalo and Ilmoniemi, 1997. The interference elimination by projection was found to be robust and relatively easy to automate. The projection operators are constructed from signal space vectors corresponding to the interfering signal-space components. For MCGs, these vectors are determined by template matching, averaging, and orthogonal construction. The projection operator application, however, redistributes fMEG signals among sensors (Hamalainen, 1995), even to sensor array regions distant from the fetal head location where the fMEG signal should normally not be present. Such signal redistribution does not affect the fMEG signal analysis because the effect of projection can be included in the forward solution. However, redistribution makes it difficult to interpret the fMEG signal maps visually. Because such visual interpretation is often useful, we have devised a procedure for correction of the redistributed fMEG signal topography Vrba et al., 2002, Vrba et al., 2003.

It is possible to correct projection redistribution if one assumes that there exists a set of channels in which the fMEG signal should be small enough to be negligible Vrba et al., 2002, Vrba et al., 2003. The absence of the fMEG signal is assured only in channels distant from the known fetal head position. Therefore, this approach will work well only if the fMEG sensing array covers a large area of the maternal abdomen and the number of channels is large. Such correction does not require any knowledge of the fetal brain model and is general and applicable to all multichannel measurements where the signal is redistributed by application of projection operators.

We also demonstrate an alternate correction method which may be used in cases where the fMEG signal can be described by a single equivalent current dipole (ECD), which is a well-established approach in adult MEG especially in the case of primary sensory processes Grynszpan and Geselowitz, 1973, Hamalainen et al., 1993, Lounasmaa et al., 1996, when the assumption of a single active source is valid. The redistribution is incorporated in the forward solution, the redistributed fMEG signal is fitted by an ECD, and the dipole forward solution is used in lieu of the corrected fMEG signal. In the case of fetal brain activity, the present knowledge about regional specialization and the current flow during cortical activation is rather limited. We assume that a single ECD is an appropriate model for description of the recorded activity. This assumption is reasonably satisfied because the volume of the brain during the last trimester is around 300–500 cm3 (Endres and Cohen, 2001), compared to a mean of l400 cm3 for adults, and the fetal head is typically more distant from the sensor array. Such an ECD model then describes the mean effect of all simultaneously activated areas.

The amplitudes of the evoked fMEG signals are comparable to that of the averaged background of other signals and could thus be confused with artifacts. Also, it is difficult to judge whether the redistribution was restored correctly. To assure consistency, we must ask whether the measured fMEG and the subsequent corrected signal are physiologically meaningful and whether both can be explained by sources within the fetal head. We propose that one necessary step to answering these questions is to test whether the fMEG signals are consistent with fetal brain models. We investigate the reliability of the fMEG signals and the projection operator restoration by using two extreme fetal brain models (Vrba et al., 2002): uniform abdomen model and fetal head model. We show that the measured fMEG signals are consistent with the hypothesis that they originate within the fetal head and that after correction for the projection redistribution they satisfy the same model as before the correction. It was also found that the fetal head model is preferred because it accounts more accurately for the observed signals than the uniform abdomen model.

Further corroboration of the fMEG signal authenticity is obtained in situations where different spontaneous and evoked signals are evaluated from the same dataset. We describe a measurement where during the collection of visual-evoked fMEG, we have also detected sharp waves and spontaneous fMEG. All three signal types were corrected for projection operator redistribution, and it was shown that each can be explained well by a single source using the fetal head model. In addition, the fetal head positions obtained by modeling these three signal types were in the same region of the maternal abdomen and in good correspondence with the fetal head position determined by ultrasound.

Section snippets

Experimental

Measurements were performed with a 151-channel SQUID fMEG system (SARA) (Robinson et al., 2001). SARA is a stationary, floor-mounted instrument where the mother sits and leans her abdomen against an anatomically shaped sensing surface. This design is inherently safe. The mother is comfortable and can gain easy access to or dismount from the system. The system is in a magnetically shielded room (MSR) and is equipped with high-order synthetic gradiometer noise cancellation which effectively

Results and verification of fetal brain signals

Artifact elimination by orthogonal projection causes redistribution of the resulting fMEG signal (Hamalainen, 1995), and, in situations where the fMEG and MCG vectors are not orthogonal, can cause reduction of fMEG amplitude. The redistribution is accounted for in the forward solution, and if required, can also be corrected Vrba et al., 2002, Vrba et al., 2003. The correction accuracy and the demonstration that the fMEG signal originates within the fetal head are analyzed in this section by

Conclusions

Fetal brain MEG measurements are contaminated by magnetic interference originating primarily from maternal and fetal hearts. This interference can be removed by determining the characteristic MCG signal space vectors and then projecting the mMCG and fMCG out. In the case of flash-evoked fMEG, the resulting signal without the MCG artifacts is averaged. In the case of spontaneous and sharp wave fMEG, the signal amplitudes are large enough and can be observed without averaging.

Removal of mMCG and

Acknowledgements

This study was partially supported by grants from the National Institutes of Health, 1R33 EB00978-01 (NIBIB) and 5R01 NS36277-03 (NINDS), USA. We would also like to acknowledge the support of CTF Systems Inc., a subsidiary of VSM MedTech Ltd., during the final phases of the work.

References (35)

  • S.N. Erne et al.

    Magnetocardiography, an introduction

  • M.S. Hamalainen

    Functional localization based on measurements with a whole-head magnetometers system

    Brain Topogr.

    (1995)
  • M. Hamalainen et al.

    Magnetoencephalography—Theory, instrumentation, and applications to noninvasive studies of working human brain

    Rev. Mod. Phys.

    (1993)
  • J.H. Holland

    Adaptation in Natural and Artificial Systems

    (1975)
  • M. Huotilainen et al.

    The projection method in removing eye-blink artifacts from multichannel MEG measurements

  • K. Kirkpatric et al.

    Optimization by simulated annealing

    Science

    (1983)
  • O.V. Lounasmaa et al.

    Information processing in the human brain: magnetoencephalographic approach

    Proc. Natl. Acad. Sci. U. S. A.

    (1996)
  • Cited by (46)

    • Verification of fetal evoked response by magnetic dipole fitting

      2021, Neuroscience Letters
      Citation Excerpt :

      Most of the studies that experimented with dipole fitting to fMEG data focused on ECD. Vrba et al. [18] used a spherical head model, where the fetal head was assumed to be electrically isolated from the maternal abdomen and currents flow within the fetal head volume. By fitting an ECD, they have verified that the recorded signal originated from fetal head.

    • Toward noninvasive monitoring of ongoing electrical activity of human uterus and fetal heart and brain

      2017, Clinical Neurophysiology
      Citation Excerpt :

      Fig. 2, left, shows an example of the partial-coverage array with a 120° coverage consisting of 5 × 8 sensors, 40 sensors total, on the anterior side of the torso of the pregnant woman. This array covers the abdominal surface similar to the 151-channel biomagnetometer called SARA (SQUID array for reproductive assessment) (Vrba et al., 2004), which is the most advanced FM scanner in use today. Although elegantly designed SARA provides a limited coverage over the abdominal surface.

    • Magnetomyography of the levator muscle complex: a novel assessment tool

      2016, American Journal of Obstetrics and Gynecology
    • Functional plasticity before the cradle: A review of neural functional imaging in the human fetus

      2013, Neuroscience and Biobehavioral Reviews
      Citation Excerpt :

      These post hoc corrections have proven to increase reliability in signal localization. Future studies will quite certainly continue to hone these nascent procedures (McCubbin et al., 2008; Sheridan et al., 2008; Vrba et al., 2004). Another challenge involved in MEG signal detection has been properly locating where the fetal head is located in relation to signals.

    View all citing articles on Scopus
    View full text