Source characterization using linear transformations of multichannel biomagnetic recordings

• Κύριος συγγραφέας: Popescu, Elena-Anda
• Άλλοι συγγραφείς: Κωστόπουλος, Γεώργιος
• Μορφή: Thesis
• Γλώσσα: English
• Έκδοση: 2015
• Θέματα: Neuroscience; Magnetoencephalography; Brain tomography; Neuroinformatics; Νευροεπιστήμες; Νευροπληροφορική; Μαγνητοεγκεφαλογραφία; Τομογραφία εγκεφάλου; 616.075 48

The biomagnetic measurements offer completely non-invasive and non-contact multichannel recordings with millisecond temporal resolution that provide the direct assessment of the functional dynamics of the brain, via magnetoencephalography (MEG), and heart, via magnetocardiography (MCG). Dense-array biomagnetometers have been recently designed for the investigation of the neurophysiological development of the fetus in utero. Fetal MCG offers accurate recordings of the fetal cardiac activity that allow to examine the effects of different maternal factors (such as smoking, exercise, fatty-acids diet supplements) on the fetal neurophysiological development using measures of heart rate and heart rate variability, and high-resolution measurement of the cardiac time intervals. However, the extraction of the fetal MCG signal relies on the identification and elimination of other signal contributions associated with the maternal cardiac activity or specific fetal behavioral patterns, such as fetal sucking and hiccup activity. The results of the present thesis demonstrate the potential of the linear transformations to generate data representations that enable straightforward and effective extraction and characterization of the biomagnetic signal components, and the elimination of the interference. The independent component analysis has been proven as an efficient signal separation and filtering technique for fetal biomagnetic recordings. ICA has been shown to provide adequate interference elimination of the maternal and fetal cardiac activity for the separation of the signal associated with distinct fetal behavioral patterns. Specifically, ICA has been applied to the extraction of the non-nutritive sucking pattern in the fetus. The recorded biomagnetic signal was characterized by the distinctive rhythmic pattern of bursts documented in the fetus using Doppler ultrasonography. The signal characteristics were consistent with the pacifier-induced non-nutritive sucking recorded in the infant using measurements of the sucking pressure variation and electromyography (EMG)recordings. Additionally, the measurements in fetuses have been confirmed by the biomagnetic recordings of the pacifier-induced non-nutritive sucking in infants. Correspondingly, the signal pattern associated with the fetal hiccup activity has been recorded and characterized using biomagnetic measurements in the fetus. The signal waveform was consistent with the adult EMG data of the diaphragm motor response to electrical or magnetic stimulation of the phrenic nerve. Additionally, the simultaneous recording of the fetal cardiac activity allows the estimation of the fetal heart rate signal. Thus, it has been possible to assess the interaction between the fetal behavior and the fetal heart rhythm. Although the linear transformations of the biomagnetic data, such as the signal subspace separation, beamforming or independent component analysis methods, do not provide direct information on the location of the sources, they generate data representations that enable the separation of the distinct signal components. The signal subspace separation method based on irreducible tensor representation approach may potentially allow to segregate and remove the contribution of the external interference sources, e.g. an active ultrasound device for real-time monitoring during fetal biomagnetic recording. The preliminary assessment indicates the ability of the irreducible tensor representation technique to provide an effective data transformation for the compensation of fetal movement during the recording that affects the accuracy of the averaged fetal cardiac signal. Additionally, the modified beamformers with suppression region that integrate a priori anatomical information in order to perform an adaptive suppression of the interferers may be proven more efficient for accurate signal estimation of the fetal spontaneous brain activity, and auditory and visual evoked responses.