Jack McCubbin

Fetal MEG redistribution by projection operators

The fetal magnetoencephalogram (fMEG) is measured in the presence of large interference from the maternal and fetal magnetocardiograms. This interference can be efficiently attenuated by orthogonal projection of the corresponding spatial vectors. However, the projection operators redistribute the fMEG signal among sensors. Although redistribution can be readily accounted for in the forward solution, visual interpretation of the fMEG signal topography is made difficult. We have devised a general, model-independent method for correction of the redistribution effect that utilizes the assumption that we know in which channels the fMEG should be negligible (such channels are distant from the known fetal head position). In a simplified case where the fMEG can be explained by equivalent current dipoles, the correction can also be obtained from fitting the dipoles to the fMEG signal. The corrected fMEG signal topography then corresponds to the dipole forward solution, but without orthogonal projection. We illustrate the redistribution correction on an example of experimentally measured flash evoked fMEG.

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

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.

Bootstrap significance of low SNR evoked response

In order to obtain adequate signal to noise ratio (SNR), stimulus-evoked brain signals are averaged over a large number of trials. However, in certain applications, e.g. fetal magnetoencephalography (MEG), this approach fails due to underlying conditions (inherently small signals, non-stationary/poorly characterized signals, or limited number of trials). The resulting low SNR makes it difficult to reliably identify a response by visual examination of the averaged time course, even after pre-processing to attenuate interference. The purpose of this work was to devise an intuitive statistical significance test for low SNR situations, based on non-parametric bootstrap resampling. We compared a two-parameter measure of p-value and statistical power with a bootstrap equal means test and a traditional rank test using fetal MEG data collected with a light flash stimulus. We found that the two-parameter measure generally agreed with established measures, while p-value alone was overly optimistic. In an extension of our approach, we compared methods to estimate the background noise. A method based on surrogate averages resulted in the most robust estimate. In summary we have developed a flexible and intuitively satisfying bootstrap-based significance measure incorporating appropriate noise estimation.

Validation of the flash-evoked response from fetal MEG.

Flash-evoked responses can be recorded from the fetus in utero. However, a standard analysis approach based on orthogonal projection (OP) to attenuate maternal and fetal cardiac signals leads to a spatial redistribution of the signal. This effect prevents the correlation of source location with a known fetal head location in some cases and the signal-to-noise ratio (SNR) is sometimes limited such that the response latency is difficult to determine. We used a modified beamformer model search analysis to avoid the redistribution shortcoming and to improve the SNR. We included a statistical test for residual interference in the average and quantified significance of the evoked response with a bootstrap method. Selected source locations compared favorably to fetal head locations estimated from ultrasound exams. The evoked response time course was found to have a significant post-trigger peak with a latency between about 180 and 770 ms in more than 90% of the subject measurements. These results confirm that the combined application of a beamformer model search and bootstrap significance test provides a validation of the flash-evoked response observed in OP processed fetal MEG channels.

Removal of interference from fetal MEG by frequency dependent subtraction

Fetal magnetoencephalography (fMEG) recordings are contaminated by maternal and fetal magnetocardiography (MCG) signals and by other biological and environmental interference. Currently, all methods for the attenuation of these signals are based on a time-domain approach. We have developed and tested a frequency dependent procedure for removal of MCG and other interference from the fMEG recordings. The method uses a set of reference channels and performs subtraction of interference in the frequency domain (SUBTR). The interference-free frequency domain signals are converted back to the time domain. We compare the performance of the frequency dependent approach with our present approach for MCG attenuation based on orthogonal projection (OP). SUBTR has an advantage over OP and similar template approaches because it removes not only the MCG but also other small amplitude biological interference, avoids the difficulties with inaccurate determination of the OP operator, provides more consistent and stable fMEG results, does not cause signal redistribution, and if references are selected judiciously, it does not reduce fMEG signal amplitude. SUBTR was found to perform well in simulations and on real fMEG recordings, and has a potential to improve the detection of fetal brain signals. The SUBTR removes interference without the need for a model of the individual interference sources. The method may be of interest for any sensor array noise reduction application where signal-free reference channels are available.

Searching for the best model: ambiguity of inverse solutions and application to fetal magnetoencephalography.

Fetal brain signals produce weak magnetic fields at the maternal abdominal surface. In the presence of much stronger interference these weak fetal fields are often nearly indistinguishable from noise. Our initial objective was to validate these weak fetal brain fields by demonstrating that they agree with the electromagnetic model of the fetal brain. The fetal brain model is often not known and we have attempted to fit the data to not only the brain source position, orientation and magnitude, but also to the brain model position. Simulation tests of this extended model search on fetal MEG recordings using dipole fit and beamformers revealed a region of ambiguity. The region of ambiguity consists of a family of models which are not distinguishable in the presence of noise, and which exhibit large and comparable SNR when beamformers are used. Unlike the uncertainty of a dipole fit with known model plus noise, this extended ambiguity region yields nearly identical forward solutions, and is only weakly dependent on noise. The ambiguity region is located in a plane defined by the source position, orientation, and the true model centre, and will have a diameter approximately 0.67 of the modelled fetal head diameter. Existence of the ambiguity region allows us to only state that the fetal brain fields do not contradict the electromagnetic model; we can associate them with a family of models belonging to the ambiguity region, but not with any specific model. In addition to providing a level of confidence in the fetal brain signals, the ambiguity region knowledge in combination with beamformers allows detection of undistorted temporal waveforms with improved signal-to-noise ratio, even though the source position cannot be uniquely determined.

Verification of fetal brain responses by coregistration of fetal ultrasound and fetal magnetoencephalography data

Fetal magnetoencephalography (fMEG) is used to study neurological functions of the developing fetus by measuring magnetic signals generated by electrical sources within the fetal brain. For this aim either auditory or visual stimuli are presented and evoked brain activity or spontaneous activity is measured at the sensor level. However a limiting factor of this approach is the low signal to noise ratio (SNR) of recorded signals. To overcome this limitation, advanced signal processing techniques such as spatial filters (e.g., beamformer) can be used to increase SNR. One crucial aspect of this technique is the forward model and, in general, a simple spherical head model is used. This head model is an integral part of a model search approach to analyze the data due to the lack of exact knowledge about the location of the fetal head. In the present report we overcome this limitation by a coregistration of volumetric ultrasound images with fMEG data. In a first step we validated the ultrasound to fMEG coregistration with a phantom and were able to show that the coregistration error is below 2 cm. In the second step we compared the results gained by the model search approach to the exact location of the fetal head determined on pregnant mothers by ultrasound. The results of this study clearly show that the results of the model search approach are in accordance with the location of the fetal head.

Fetal MEG evoked response latency from beamformer with random field theory

Analysis of fetal magnetoencephalographic brain recordings is restricted by low signal to noise ratio (SNR) and non-stationarity of the sources. Beamformer techniques have been applied to improve SNR of fetal evoked responses. However, until now the effect of non-stationarity was not taken into account in detail, because the detection of evoked responses is in most cases determined by averaging a large number of trials. We applied a windowing technique to improve the stationarity of the data by using short time segments recorded during a flash-evoked study. In addition, we implemented a random field theory approach for more stringent control of false-positives in the statistical parametric map of the search volume for the beamformer. The search volume was based on detailed individual fetal/maternal biometrics from ultrasound scans and fetal heart localization. Average power over a sliding window within the averaged evoked response against a randomized average background power was used as the test z-statistic. The significance threshold was set at 10% over all members of a contiguous cluster of voxels. There was at least one significant response for 62% of fetal and 95% of newborn recordings with gestational age (GA) between 28 and 45 weeks from 29 subjects. We found that the latency was either substantially unchanged or decreased with increasing GA for most subjects, with a nominal rate of about -11 ms/week. These findings support the anticipated neurophysiological development, provide validation for the beamformer model search as a methodology, and may lead to a clinical test for fetal cognitive development.