- Email: [email protected]
Detection of fetal auditory evoked responses by means of magnetoencephalography
Brain Research 917 (2001) 167–173 www.elsevier.com / locate / bres
Research report
Detection of fetal auditory evoked responses by means of magnetoencephalography a a,b , c d,b Filippo Zappasodi , Franca Tecchio *, Vittorio Pizzella , Emanuele Cassetta , d e f,b,d Giuseppe V. Romano , Giancarlo Filligoi , Paolo M. Rossini b
a Istituto di Elettronica dello Stato Solido ( IESS), C.N.R., Roma, Italy I.R.C.C.S. ‘ Centro S.Giovanni di Dio’, c /o Ist. Sacro Cuore di Gesu` , via Pilastroni, Brescia, Italy c ITAB and Istituto di Fisica Medica, Universita` ‘ G.d’ Annunzio’, Chieti, Italy d AFaR., Dip. Neuroscienze, Dip. Ginecologia Osp.Fatebenefratelli, Isola Tiberina, Roma, Italy e Dip. INFOCOM, Universita` ‘‘ La Sapienza’’, Roma, Italy f Neurologia, Universita` Campus Biomedico, Roma, Italy
Accepted 24 July 2001
Abstract MagnetoEncephaloGraphy (MEG) is proposed as a non-invasive technique to detect the physiological activity of fetal brain, due to its ability to record brain activity without direct contact with the head and the transparency of magnetic signals in passing through extracerebral fetal layers and the mother’s abdomen. Healthy women with uncomplicated pregnancies and fetuses in breech presentation were examined; gestational ages at time of study ranged between 36 and 40 weeks. In order to evaluate fetal well-being, ultrasound and cardiotocographic data were assessed a few days before and after MEG recording sessions. The participating women were placed in a semi-reclining position in a magnetically shielded room; here the presentation of the fetus and precise region of the mother’s abdomen corresponding to the fetal head were determined by ultrasound investigation in order to place the MEG detecting system as near as possible to the fetal brain. MEG recordings were performed by means of a 28-channel neuromagnetic system. Every MEG recording session was performed during the acoustic stimulation of fetuses, in order to detect the cerebral events evoked by peripheral stimuli. The auditory stimuli were delivered from a plastic tube placed on mother’s abdomen, near the fetal head, and consisted of a 300 ms 103 dB pure tone at 500 and 1000 Hz, presented at a 0.4 c / s repetition rate. In six cases following accurate digital subtraction of maternal and fetal electrocardiographic (EKG) signals we remained with a stimulus-related response peaking at about 250 ms; this was considered to originate from the fetal brain. In favour of this in three cases a clear dipolar distribution was evident at the peak of brain response centered on the fetal head and consistent with a brain generator. Despite several technical problems requiring solution before a possible routine clinical application, MEG has been found to be suitable for the non-invasive exploration of the fetal brain. 2001 Elsevier Science B.V. All rights reserved. Theme: Sensory systems Topic: Auditory systems: central physiology Keywords: Magnetoencephalography; Brain development; Fetus cerebral activity; Auditory system
1. Introduction The development and shaping of neuronal connections largely depend upon the experience-related brain cells firing during different stages of fetal life. This is particular
*Corresponding author. Tel.: 139-6-683-7382; fax: 139-6-683-7385. E-mail address: [email protected] (F. Tecchio).
important in the developing sensory pathways [19,22]. Both spontaneous and stimulus related electromagnetic brain activity is very important for the correct constitution of inter-neuronal contact as it is the basis of the correct arrangement of the nervous circuits and pathways of all the subsystems of the nervous central system. The complexity of the nervous circuits reflects both the expression of the genetic program and the outcome of use-dependent plasticity stemming from environmental factors, in particular
0006-8993 / 01 / $ – see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0006-8993( 01 )02901-8
168
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
relating to epochs corresponding to the last phases of gestation and the early months of postnatal life [7,16]. The sense of hearing is already functioning in some way in the last 2 months of intrauterine life, as has been demonstrated by fetal motor responses and heart rate variations evoked by sounds; as a consequence, hearing is the neurosensorial activity most studied in both animal and human fetuses, mainly through indirect techniques exploring behavioral responses to acoustic stimulation [8,9,12,11,4]. In the early ‘90s psychological evaluations were undertaken in order to investigate whether newborns recognize the sounds heard during intrauterine life; it was demonstrated that human newborns pay more attention to the mother’s voice than to any other human voice and that they do not show alert reactions to alarming acoustic stimuli if they were accustomed having been frequently administered during fetal life [10,12,3]. More recently systematic studies have been undertaken on fetuses of different gestational ages in order to gain insight into intrauterine developmental changes of the hearing system, by means of the indirect evaluation of motor activity [6] or heart rate modifications [8] induced by acoustic stimuli at different frequencies. The tonal frequency at which a response can be observed is initially around 500 Hz at 19 weeks of gestational age. Later on, sensitivity to the range of lower frequency (100–250 Hz) and subsequently of high frequency tones (1000–3000 Hz) is developed. It is generally accepted that healthy fetuses would respond to a 1000 Hz stimulus at 33–35 weeks of gestational age. The acquisition of linguistic competency may well be traced to intrauterine experience; that is, nervous pathways would acquire the ability to discriminate phonemes on the basis of the acoustic stimuli perceived during fetal life [15]. Despite recent advances in functional brain imaging technology, there are still no suitable methods for the analysis of fetal cerebral activity. MEG is proposed as a non-invasive technique for the detection of fetal brain activity; firstly, because magnetic signals transparently pass through different extracerebral and other body tissue layers and, secondly, no direct contact between the recording device and the biological signal source is required. In order to assess fetal cerebral activity in the perspective of these findings, the first step of our project focused on the extraction of evoked responses by auditory stimulation; the results are detailed in the present study.
2. Materials and methods Twelve pregnant females were enrolled among the obstetric clinic outpatients of the Fatebenefratelli Hospital in Rome. Informed consent to the study was obtained. The research was approved by the local ethical committee.
Inclusion criteria comprised the following: (1) gestational age.35 weeks; (2) uncomplicated pregnancy without any known fetal distress; (3) fetal well-being demonstrated by standard instrumental examination; (4) fetal weight. 2500 g; (5) transverse or breech presentation.
2.1. Clinical evaluation All enrolled subjects underwent the following evaluations: • Ultrasound fetal analysis aimed at confirming gestational age, excluding brain malformations and estimating fetal weight and volume of amniotic fluid. • Echo-fluximetry of maternal and fetal districts aimed at evaluating the oxygen exchanges between mother and fetus. Resistance in the umbilical arteries, fetal middle cerebral artery and / or fetal abdominal aorta were evaluated. • Cardio-tocographic monitoring, performed for a period of 25–30 min, every fourth day until the delivery. Recording of fetal activity also indirectly measures fetal central nervous system integrity and fetal wellbeing in utero (Fig. 1b).
2.2. MEG recordings 2.2.1. Procedure • Ecographic assessment of the precise location of the fetal head position was performed immediately before MEG sensor positioning. The distance of fetal head from the device was also calculated echographically. • MEG recording device was positioned under ultrasound guidance upon the abdominal region corresponding to the fetal head. • The magnetic field distribution in the region above the fetal head was measured by the 28-channel system operating in a magnetically shielded room (Vacuumschmelze GMBH) at IESS-CNR in Rome. This system features 16 first-order axial gradiometers and 9 magnetometers, as well as 3 balancing magnetometers for noise cancellation coupled with low noise dcSQUIDs, with an overall sensitivity of about 5–7 fT / Hz 1 / 2 . Gradiometeres and magnetometers are characterized by a noise of 5–6 fT / Hz 1 / 2 and 7–9 fT / Hz 1 / 2 respectively. The 25 measuring sites were regularly distributed on a spherical surface with a 135 mm radius of curvature, covering an area of about 180 cm 2 . A separate electrical channel was used to record the electrocardiogram (EKG) of the mother. • Auditory stimuli consisted of pure tone bursts with carrier frequency of 500 and 1000 Hz, presented every 2631 ms, lasting 500 ms, externally delivered with an
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
169
Fig. 1. (a) Position of the MEG system on the maternal abdomen region nearest to the fetus head, as indicated ecographic assessment. (b) Antenatal surveillance: Cardiotocography (GTG) is used to assess fetal well-being, and records simultaneous fetal heart rate parameters (FHRP), fetal body movements and uterine activity. Sample showing baseline FHR on upper channel (1); blocks represent cluster of fetal movements (2). On bottom line uterine contractions are detected by external tocodinamometry (3).
intensity of 103 dB SPL by a GRASS stimulator and transmitted by a plastic tube positioned over maternal abdomen, just beneath the fetal head. • In order to evaluate possible contribution from noncerebral, biological sources adjunctive recordings from abdomen positions away from the fetus head (around 20 cm) were performed in 2 cases.
2.2.2. Data analysis • Ten minutes of MEG activity from each position were continuously acquired (0.48–64 Hz bandwidth, 250 Hz of sampling rate) during auditory stimulation. The magnetic fields generated near the maternal and fetal hearts are very strong, about one-two orders of magnitude greater than the field generated by the fetal brain. Subtraction of cardiac interference was obtained by orthogonal projection [14] or adaptive algorithms [21,18]
in each trial recording. In consideration of the morphological repetition of cardiac waves and the linear independence of fetal cerebral and heart sources, an average cardiac reference was built by averaging the single magnetic channel data synchronously to R-peak of the QRS wave. Determination of occurrence of QRS complexes for triggering the averages has been made on the additional electric channels using simple threshold among the methods available [1]. The cardiac artifact was eliminated for each localized QRS-complex through orthogonal signal projection algorithm (OSPA) between the segment of the magnetic channel synchronized to the corresponding Rwave and the average heart interference. The elimination was performed twice for maternal and fetal interference. Alternatively, the more classical adaptive noise cancellation (ANC, 21) was applied on each channel, using as reference input the average interference repeated every instant of occurrence of the R-waves, since cardiac signals are characteristically nearly periodic over the normalized period length (distance R–R). However, adaptive filtering
170
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
is much more laborious than the first method because of the best choice of the convergence parameter, which generally changes for each channel and must be empirically set. Furthermore, auditory stimulus-synchronous averaging eliminates any minimal difference between the two methods, making OSPA an easier and faster application. • Acquired data were off-line digitally forward and time-reversed filtered between 2.5 and 20 Hz. • Of the total 230 delivered auditory stimuli, around 200 artifact free trials were averaged to achieve auditory evoked fields (AEFs). The amplitude of AEFs was calculated for each channel with respect to a baseline level chosen as the mean value of the (250, 20) ms post-stimulus epoch. The results obtained on fetuses were compared with the evoked acoustic potentials (AEPs) obtained in preterm newborns in two cases (pure tone bursts at 1000 Hz delivered to the right ear, 110 dB of intensity, 500 ms duration, 2500 ms interstimulus interval, 0.1–50 Hz bandpass filters, montage Cz-A2).
3. Results The fetal scalp was typically 1.5–3.5 cm from surface of maternal abdomen. The average distance between fetal head and the MEG sensors was determined ecographically to be 3562 mm. All the pregnant women reported that the MEG recording session was comfortable, and none reported any shortor long-term discomfort. MEG signals acquired with the sensors on the abdomen region distant from the fetus head did not show any identifiable response during auditory stimulation. In six of the twelve tested subjects no reliable patterns of stimulus-related brain activity were recorded. Fig. 2 shows the procedure for heart artifact elimination on a single channel. The quality of interference elimination was tested on an artificial signal simulating the real recordings (Fig. 2 left), obtained by superimposing three different magnetic patterns of data: a weighted MEG recording of an adult subjected to the same auditory stimulation described here for the fetus, a weighted magnetocardiographic (MKG) recording of the same subject simulating the maternal heart, and a weighted MKG recording rescaled in time to have approximately a double frequency compared with the previous one simulating the fetal heart. In three cases an evident deflection peaking around 200–300 ms from stimulus onset (Fig. 3a, c and e), with phase inversion among channels (Fig. 3b, d and f) was observed. In the remaining 3 cases a deflection between 200 and 300 ms was also recognizable, although the fetal-sensors relative position was not suitable to map out a
dipolar distribution. The amplitude of these evoked responses was very low, about 50–80 fT as maximal– minimal channels difference at peak latency. The small number of identifiable AEF did not permit analysis of the differences between responses to 500 and 1000 Hz stimulations. AEP recordings on the pre-term newborns showed a clear deflection around 200 ms in both babies (38 weeks of gestational age, Fig. 4a, b).
4. Discussion In different preliminary recording sessions we observed that magnetic fetal signals are favorably detected when the fetus is in breech presentation. In fact, this fetal presentation allows the closest positioning of the recording system to the fetal head. On the other hand, however, this position is also contaminated by the strongest signal from the maternal heart which overlaps with the fetal brain activity, highlighting the crucial importance of the filtering operation. Our results on test signals demonstrated that our method for eliminating EKG artifacts is quite efficient (Fig. 2 left); moreover, two adjunctive methods could be applied in real fetal MEG (fMEG), namely the use of three EKG channels for maternal heart artifact extraction, and filtering techniques not averaging MKG but directly eliminating the contaminated trials. Along this line, a filtering operation using PCA and ICA could be developed [23,2]. Hearing activity of fetuses is lower than that of newborns of comparable gestational age. In fact, sound is transmitted from the maternal abdomen to the fetal head through the amniotic fluid, where the basal noise level is about 15 dB for the frequencies to which the auditory system responds and the cardiac maternal activity and vascular sound are 25 dB stronger than the basal noise [13]. The amniotic fluid attenuates frequencies below 500 Hz less than 5 dB and higher frequencies up to 30 dB. Furthermore, fetal hearing system is stimulated not through the external and middle ear, but through bone conduction, and consequently sound energy is reduced during the passage through the bone of the skull. This reduction ranges from 10 up to 20 dB below 250 Hz and 40 up to 50 dB at frequencies from 500 up to 2000 Hz for fetus scalp [4,5]. In this regard, the experimental set-up would probably require a larger contact surface at the maternal abdomen (around 1–2 dm 2 ). In addition, lower content of oxygen during placentar exchanges compared with that during pulmonary exchanges in extrauterine life induces fetal hypo-acousia, because the magnitude of the endocochlear potential is dependent on oxygen supply [17]. In half of the examined fetuses, wave responses to acoustic stimuli were reliably recorded [20]. They disappeared if the MEG sensor was shifted on the mother’s abdomen to key it away from the optimal position overly-
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
171
Fig. 2. (Left) Traces of one MEG channel used to create the test signal (a–d) for the filter evaluation (e to h): (a) cerebral adult signal on temporal region during auditory stimulation, (b) MKG signal, (c) time-rescaled MKG, (d) weighted sum of the preceding signals; (e) mean simulation of ‘maternal MKG’ (repeated every instant of occurrence of R waves), (f) trace ‘d’ after simulated ‘maternal’ MKG elimination, (g) mean ‘fetal’ MKG, (h) final OSPA-filtered signal, to be compared for filter evaluation with trace a (repeated lastly). (Right) Result of cardiac interference OSPA elimination in a single channel of fMEG recording: (a) recorded fMEG signal, (b) mean maternal heart artifact, (c) result of maternal interference elimination, (d) mean fetal heart artifact (e) signal OSPA-filtered.
ing the fetal brain; moreover, in some cases a dipolar distribution corresponding with the abdominal projection of the fetal head was found. These findings suggest that such responses are cerebral in origin. The present study represents a very first step in the systematic assessment of fetal cerebral activity. The improvement of non-invasive methods to record spontaneous and stimulus-related electromagnetic activity of the fetal
brain would be of paramount importance, as present techniques do not yet enable to reliably record fetal brain activity. MEG devices recently developed ad hoc for fetal recordings could more efficiently monitor the extraction of both heart and brain signals. This method, when combined with current behaviour, indirect techniques for testing fetal activity, may help in gathering prenatal information on brain functionality, both in the healthy and the diseased.
172
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
Fig. 3. Superimposition of the internal gradiometer signals during the first 400 ms from the stimulus onset, in the 3 cases (a, c, e) with identifiable phase reversal appearing at around 220 ms latencies (black arrows). Isofield maps at these latencies show clear dipolar field distributions (b, d, f). Isofield contours are stepped by 6, 8, 14 fT respectively. All cases were stimulated at 1000 Hz frequency.
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
Fig. 4. AEP recordings (Cz derivation with bi-auricular reference, two repetition for each subject) of 2 preterm newborns, 38 weeks of gestational age. Traces are replicated twice to show the stability of neonatal peaks.
References [1] S. Azevedo, R.L. Longhini, Abdominal-lead fetal electrocardiographic R-waves enhancement, IEEE Trans. Biom. Eng. 27 (1980) 255–260. [2] A.J. Bell, T.J. Sejnowski, An information maximisation approach to blind separation and blind deconvolution, Neural Comput. 7 (1995) 1129–1159.
173
[3] S.M. Damstra-Wijmenga, The memory of the new-born baby, Midwives Chron. 104 (1991) 66–69. [4] K.J. Gerhardt, R.M. Abrams, Fetal hearing: characterization of the stimulus and response, Semin. Perinatol. 20 (1996) 11–20. [5] K.J. Gerhardt, X. Huang, K.E. Arrington, K. Meixner, R.M. Abrams, P.J. Antonelli, Fetal sheep in utero hear through bone conduction, Am. J. Otolaryngol. 17 (1996) 374–379. [6] P.G. Hepper, B.S. Shahidullah, Development of fetal hearing, Arch. Dis. Child. 71 (1994) F81–F87. [7] D.H. Hubel, T.N. Wiesel, The period of susceptibility to the physiological effects of unilateral eye closure in kittens, J. Physiol. 206 (1970) 419–436. [8] B. Johansson, E. Wedenberg, B. Westin, Fetal heart rate response to acoustic stimulation in relation to fetal development and hearing impairment, Acta Obstet. Gynecol. Scand. 71 (1992) 610–615. [9] N.P. Luz, Auditory evoked response of the human fetus: simplified methodology, J. Perinat. Med. 19 (1991) 177–183. [10] E.M. Ockleford, M.A. Vince, C. Layton, M.R. Reader, Responses of neonates to parents’ and others’ voices, Early Hum. Dev. 18 (1988) 27–36. [11] M.J. Parkes, P.J. Moore, D.R. Moore, N.M. Fisk, M.A. Hanson, Behavioral changes in fetal sheep caused by vibroacoustic stimulation: the effects of cochlear ablation, Am. J. Obstet. Gynecol. 164 (1991) 1336–1343. [12] D. Querleu, X. Renard, F. Versyp, L. Paris-Delrue, G. Crepin, Fetal hearing, Eur. J. Obstet. Gynecol. Reprod. Biol. 28 (1988) 191–212. [13] D. Querleu, R. Xavier, C. Boutteville, G. Crepin, Hearing by the human fetus?, Semin. Perinatol. 13 (1989) 409–420. [14] M. Samonas, M. Petrou, A.A. Ioannides, Identification and elimination of cardiac contribution in single-trial magnetoencephalographic signals, IEEE Trans. Biom. Eng. 44 (1997) 386–393. [15] S. Shahidullah, P.G. Hepper, Frequency discrimination by the fetus, Early Hum. Dev. 36 (1994) 13–26. [16] Z.D. Smith, L. Gray, E.W. Rubel, Afferent influences on brainstem auditory nuclei of the chicken: n. laminaris dendritic length following monaural conductive hearing loss, J. Comp. Neurol. 220 (1983) 199–205. [17] H. Sohmer, S. Freeman, Functional development of auditory sensitivity in the fetus and neonate, J. Basic Clin. Physiol. Pharmacol. 6 (1995) 95–108. ¨ [18] P. Strobach, K. Abraham-Fuchs, W. Harer, Event synchronous cancellation of the heart interference in biomedical signals, IEEE Trans. Biom. Eng. 41 (1994) 343–350. [19] M. Sur, A. Angelucci, J. Sharma, Rewiring cortex: the role of patterned activity in development and plasticity of neocortical circuits, J. Neurobiol. 41 (1999) 33–43. [20] R.T. Wakai, C. Arthur, M.S. Leuthold, B.M. Chester, Fetal auditory evoked responses detected by magnetoencephalography, Am. J. Obstet. Gynecol. 174 (1996) 1484–1486. [21] B. Widrow et al., Adaptive Noise Canceling: Principles and Applications, Proc. IEEE 63 (1975) 1692–1716. [22] R.O. Wong, The role of spatio-temporal firing patterns in neuronal development of sensory systems, Curr. Opin. Neurobiol. 3 (1993) 595–601. [23] V. Zarzoso, A.K. Nandi, E. Bacharakis, Maternal and foetal ECG separation using blind source separation methods, IMA J. Math. Appl. Med. Biol. 14 (1997) 207–225.
Research report
Detection of fetal auditory evoked responses by means of magnetoencephalography a a,b , c d,b Filippo Zappasodi , Franca Tecchio *, Vittorio Pizzella , Emanuele Cassetta , d e f,b,d Giuseppe V. Romano , Giancarlo Filligoi , Paolo M. Rossini b
a Istituto di Elettronica dello Stato Solido ( IESS), C.N.R., Roma, Italy I.R.C.C.S. ‘ Centro S.Giovanni di Dio’, c /o Ist. Sacro Cuore di Gesu` , via Pilastroni, Brescia, Italy c ITAB and Istituto di Fisica Medica, Universita` ‘ G.d’ Annunzio’, Chieti, Italy d AFaR., Dip. Neuroscienze, Dip. Ginecologia Osp.Fatebenefratelli, Isola Tiberina, Roma, Italy e Dip. INFOCOM, Universita` ‘‘ La Sapienza’’, Roma, Italy f Neurologia, Universita` Campus Biomedico, Roma, Italy
Accepted 24 July 2001
Abstract MagnetoEncephaloGraphy (MEG) is proposed as a non-invasive technique to detect the physiological activity of fetal brain, due to its ability to record brain activity without direct contact with the head and the transparency of magnetic signals in passing through extracerebral fetal layers and the mother’s abdomen. Healthy women with uncomplicated pregnancies and fetuses in breech presentation were examined; gestational ages at time of study ranged between 36 and 40 weeks. In order to evaluate fetal well-being, ultrasound and cardiotocographic data were assessed a few days before and after MEG recording sessions. The participating women were placed in a semi-reclining position in a magnetically shielded room; here the presentation of the fetus and precise region of the mother’s abdomen corresponding to the fetal head were determined by ultrasound investigation in order to place the MEG detecting system as near as possible to the fetal brain. MEG recordings were performed by means of a 28-channel neuromagnetic system. Every MEG recording session was performed during the acoustic stimulation of fetuses, in order to detect the cerebral events evoked by peripheral stimuli. The auditory stimuli were delivered from a plastic tube placed on mother’s abdomen, near the fetal head, and consisted of a 300 ms 103 dB pure tone at 500 and 1000 Hz, presented at a 0.4 c / s repetition rate. In six cases following accurate digital subtraction of maternal and fetal electrocardiographic (EKG) signals we remained with a stimulus-related response peaking at about 250 ms; this was considered to originate from the fetal brain. In favour of this in three cases a clear dipolar distribution was evident at the peak of brain response centered on the fetal head and consistent with a brain generator. Despite several technical problems requiring solution before a possible routine clinical application, MEG has been found to be suitable for the non-invasive exploration of the fetal brain. 2001 Elsevier Science B.V. All rights reserved. Theme: Sensory systems Topic: Auditory systems: central physiology Keywords: Magnetoencephalography; Brain development; Fetus cerebral activity; Auditory system
1. Introduction The development and shaping of neuronal connections largely depend upon the experience-related brain cells firing during different stages of fetal life. This is particular
*Corresponding author. Tel.: 139-6-683-7382; fax: 139-6-683-7385. E-mail address: [email protected] (F. Tecchio).
important in the developing sensory pathways [19,22]. Both spontaneous and stimulus related electromagnetic brain activity is very important for the correct constitution of inter-neuronal contact as it is the basis of the correct arrangement of the nervous circuits and pathways of all the subsystems of the nervous central system. The complexity of the nervous circuits reflects both the expression of the genetic program and the outcome of use-dependent plasticity stemming from environmental factors, in particular
0006-8993 / 01 / $ – see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S0006-8993( 01 )02901-8
168
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
relating to epochs corresponding to the last phases of gestation and the early months of postnatal life [7,16]. The sense of hearing is already functioning in some way in the last 2 months of intrauterine life, as has been demonstrated by fetal motor responses and heart rate variations evoked by sounds; as a consequence, hearing is the neurosensorial activity most studied in both animal and human fetuses, mainly through indirect techniques exploring behavioral responses to acoustic stimulation [8,9,12,11,4]. In the early ‘90s psychological evaluations were undertaken in order to investigate whether newborns recognize the sounds heard during intrauterine life; it was demonstrated that human newborns pay more attention to the mother’s voice than to any other human voice and that they do not show alert reactions to alarming acoustic stimuli if they were accustomed having been frequently administered during fetal life [10,12,3]. More recently systematic studies have been undertaken on fetuses of different gestational ages in order to gain insight into intrauterine developmental changes of the hearing system, by means of the indirect evaluation of motor activity [6] or heart rate modifications [8] induced by acoustic stimuli at different frequencies. The tonal frequency at which a response can be observed is initially around 500 Hz at 19 weeks of gestational age. Later on, sensitivity to the range of lower frequency (100–250 Hz) and subsequently of high frequency tones (1000–3000 Hz) is developed. It is generally accepted that healthy fetuses would respond to a 1000 Hz stimulus at 33–35 weeks of gestational age. The acquisition of linguistic competency may well be traced to intrauterine experience; that is, nervous pathways would acquire the ability to discriminate phonemes on the basis of the acoustic stimuli perceived during fetal life [15]. Despite recent advances in functional brain imaging technology, there are still no suitable methods for the analysis of fetal cerebral activity. MEG is proposed as a non-invasive technique for the detection of fetal brain activity; firstly, because magnetic signals transparently pass through different extracerebral and other body tissue layers and, secondly, no direct contact between the recording device and the biological signal source is required. In order to assess fetal cerebral activity in the perspective of these findings, the first step of our project focused on the extraction of evoked responses by auditory stimulation; the results are detailed in the present study.
2. Materials and methods Twelve pregnant females were enrolled among the obstetric clinic outpatients of the Fatebenefratelli Hospital in Rome. Informed consent to the study was obtained. The research was approved by the local ethical committee.
Inclusion criteria comprised the following: (1) gestational age.35 weeks; (2) uncomplicated pregnancy without any known fetal distress; (3) fetal well-being demonstrated by standard instrumental examination; (4) fetal weight. 2500 g; (5) transverse or breech presentation.
2.1. Clinical evaluation All enrolled subjects underwent the following evaluations: • Ultrasound fetal analysis aimed at confirming gestational age, excluding brain malformations and estimating fetal weight and volume of amniotic fluid. • Echo-fluximetry of maternal and fetal districts aimed at evaluating the oxygen exchanges between mother and fetus. Resistance in the umbilical arteries, fetal middle cerebral artery and / or fetal abdominal aorta were evaluated. • Cardio-tocographic monitoring, performed for a period of 25–30 min, every fourth day until the delivery. Recording of fetal activity also indirectly measures fetal central nervous system integrity and fetal wellbeing in utero (Fig. 1b).
2.2. MEG recordings 2.2.1. Procedure • Ecographic assessment of the precise location of the fetal head position was performed immediately before MEG sensor positioning. The distance of fetal head from the device was also calculated echographically. • MEG recording device was positioned under ultrasound guidance upon the abdominal region corresponding to the fetal head. • The magnetic field distribution in the region above the fetal head was measured by the 28-channel system operating in a magnetically shielded room (Vacuumschmelze GMBH) at IESS-CNR in Rome. This system features 16 first-order axial gradiometers and 9 magnetometers, as well as 3 balancing magnetometers for noise cancellation coupled with low noise dcSQUIDs, with an overall sensitivity of about 5–7 fT / Hz 1 / 2 . Gradiometeres and magnetometers are characterized by a noise of 5–6 fT / Hz 1 / 2 and 7–9 fT / Hz 1 / 2 respectively. The 25 measuring sites were regularly distributed on a spherical surface with a 135 mm radius of curvature, covering an area of about 180 cm 2 . A separate electrical channel was used to record the electrocardiogram (EKG) of the mother. • Auditory stimuli consisted of pure tone bursts with carrier frequency of 500 and 1000 Hz, presented every 2631 ms, lasting 500 ms, externally delivered with an
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
169
Fig. 1. (a) Position of the MEG system on the maternal abdomen region nearest to the fetus head, as indicated ecographic assessment. (b) Antenatal surveillance: Cardiotocography (GTG) is used to assess fetal well-being, and records simultaneous fetal heart rate parameters (FHRP), fetal body movements and uterine activity. Sample showing baseline FHR on upper channel (1); blocks represent cluster of fetal movements (2). On bottom line uterine contractions are detected by external tocodinamometry (3).
intensity of 103 dB SPL by a GRASS stimulator and transmitted by a plastic tube positioned over maternal abdomen, just beneath the fetal head. • In order to evaluate possible contribution from noncerebral, biological sources adjunctive recordings from abdomen positions away from the fetus head (around 20 cm) were performed in 2 cases.
2.2.2. Data analysis • Ten minutes of MEG activity from each position were continuously acquired (0.48–64 Hz bandwidth, 250 Hz of sampling rate) during auditory stimulation. The magnetic fields generated near the maternal and fetal hearts are very strong, about one-two orders of magnitude greater than the field generated by the fetal brain. Subtraction of cardiac interference was obtained by orthogonal projection [14] or adaptive algorithms [21,18]
in each trial recording. In consideration of the morphological repetition of cardiac waves and the linear independence of fetal cerebral and heart sources, an average cardiac reference was built by averaging the single magnetic channel data synchronously to R-peak of the QRS wave. Determination of occurrence of QRS complexes for triggering the averages has been made on the additional electric channels using simple threshold among the methods available [1]. The cardiac artifact was eliminated for each localized QRS-complex through orthogonal signal projection algorithm (OSPA) between the segment of the magnetic channel synchronized to the corresponding Rwave and the average heart interference. The elimination was performed twice for maternal and fetal interference. Alternatively, the more classical adaptive noise cancellation (ANC, 21) was applied on each channel, using as reference input the average interference repeated every instant of occurrence of the R-waves, since cardiac signals are characteristically nearly periodic over the normalized period length (distance R–R). However, adaptive filtering
170
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
is much more laborious than the first method because of the best choice of the convergence parameter, which generally changes for each channel and must be empirically set. Furthermore, auditory stimulus-synchronous averaging eliminates any minimal difference between the two methods, making OSPA an easier and faster application. • Acquired data were off-line digitally forward and time-reversed filtered between 2.5 and 20 Hz. • Of the total 230 delivered auditory stimuli, around 200 artifact free trials were averaged to achieve auditory evoked fields (AEFs). The amplitude of AEFs was calculated for each channel with respect to a baseline level chosen as the mean value of the (250, 20) ms post-stimulus epoch. The results obtained on fetuses were compared with the evoked acoustic potentials (AEPs) obtained in preterm newborns in two cases (pure tone bursts at 1000 Hz delivered to the right ear, 110 dB of intensity, 500 ms duration, 2500 ms interstimulus interval, 0.1–50 Hz bandpass filters, montage Cz-A2).
3. Results The fetal scalp was typically 1.5–3.5 cm from surface of maternal abdomen. The average distance between fetal head and the MEG sensors was determined ecographically to be 3562 mm. All the pregnant women reported that the MEG recording session was comfortable, and none reported any shortor long-term discomfort. MEG signals acquired with the sensors on the abdomen region distant from the fetus head did not show any identifiable response during auditory stimulation. In six of the twelve tested subjects no reliable patterns of stimulus-related brain activity were recorded. Fig. 2 shows the procedure for heart artifact elimination on a single channel. The quality of interference elimination was tested on an artificial signal simulating the real recordings (Fig. 2 left), obtained by superimposing three different magnetic patterns of data: a weighted MEG recording of an adult subjected to the same auditory stimulation described here for the fetus, a weighted magnetocardiographic (MKG) recording of the same subject simulating the maternal heart, and a weighted MKG recording rescaled in time to have approximately a double frequency compared with the previous one simulating the fetal heart. In three cases an evident deflection peaking around 200–300 ms from stimulus onset (Fig. 3a, c and e), with phase inversion among channels (Fig. 3b, d and f) was observed. In the remaining 3 cases a deflection between 200 and 300 ms was also recognizable, although the fetal-sensors relative position was not suitable to map out a
dipolar distribution. The amplitude of these evoked responses was very low, about 50–80 fT as maximal– minimal channels difference at peak latency. The small number of identifiable AEF did not permit analysis of the differences between responses to 500 and 1000 Hz stimulations. AEP recordings on the pre-term newborns showed a clear deflection around 200 ms in both babies (38 weeks of gestational age, Fig. 4a, b).
4. Discussion In different preliminary recording sessions we observed that magnetic fetal signals are favorably detected when the fetus is in breech presentation. In fact, this fetal presentation allows the closest positioning of the recording system to the fetal head. On the other hand, however, this position is also contaminated by the strongest signal from the maternal heart which overlaps with the fetal brain activity, highlighting the crucial importance of the filtering operation. Our results on test signals demonstrated that our method for eliminating EKG artifacts is quite efficient (Fig. 2 left); moreover, two adjunctive methods could be applied in real fetal MEG (fMEG), namely the use of three EKG channels for maternal heart artifact extraction, and filtering techniques not averaging MKG but directly eliminating the contaminated trials. Along this line, a filtering operation using PCA and ICA could be developed [23,2]. Hearing activity of fetuses is lower than that of newborns of comparable gestational age. In fact, sound is transmitted from the maternal abdomen to the fetal head through the amniotic fluid, where the basal noise level is about 15 dB for the frequencies to which the auditory system responds and the cardiac maternal activity and vascular sound are 25 dB stronger than the basal noise [13]. The amniotic fluid attenuates frequencies below 500 Hz less than 5 dB and higher frequencies up to 30 dB. Furthermore, fetal hearing system is stimulated not through the external and middle ear, but through bone conduction, and consequently sound energy is reduced during the passage through the bone of the skull. This reduction ranges from 10 up to 20 dB below 250 Hz and 40 up to 50 dB at frequencies from 500 up to 2000 Hz for fetus scalp [4,5]. In this regard, the experimental set-up would probably require a larger contact surface at the maternal abdomen (around 1–2 dm 2 ). In addition, lower content of oxygen during placentar exchanges compared with that during pulmonary exchanges in extrauterine life induces fetal hypo-acousia, because the magnitude of the endocochlear potential is dependent on oxygen supply [17]. In half of the examined fetuses, wave responses to acoustic stimuli were reliably recorded [20]. They disappeared if the MEG sensor was shifted on the mother’s abdomen to key it away from the optimal position overly-
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
171
Fig. 2. (Left) Traces of one MEG channel used to create the test signal (a–d) for the filter evaluation (e to h): (a) cerebral adult signal on temporal region during auditory stimulation, (b) MKG signal, (c) time-rescaled MKG, (d) weighted sum of the preceding signals; (e) mean simulation of ‘maternal MKG’ (repeated every instant of occurrence of R waves), (f) trace ‘d’ after simulated ‘maternal’ MKG elimination, (g) mean ‘fetal’ MKG, (h) final OSPA-filtered signal, to be compared for filter evaluation with trace a (repeated lastly). (Right) Result of cardiac interference OSPA elimination in a single channel of fMEG recording: (a) recorded fMEG signal, (b) mean maternal heart artifact, (c) result of maternal interference elimination, (d) mean fetal heart artifact (e) signal OSPA-filtered.
ing the fetal brain; moreover, in some cases a dipolar distribution corresponding with the abdominal projection of the fetal head was found. These findings suggest that such responses are cerebral in origin. The present study represents a very first step in the systematic assessment of fetal cerebral activity. The improvement of non-invasive methods to record spontaneous and stimulus-related electromagnetic activity of the fetal
brain would be of paramount importance, as present techniques do not yet enable to reliably record fetal brain activity. MEG devices recently developed ad hoc for fetal recordings could more efficiently monitor the extraction of both heart and brain signals. This method, when combined with current behaviour, indirect techniques for testing fetal activity, may help in gathering prenatal information on brain functionality, both in the healthy and the diseased.
172
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
Fig. 3. Superimposition of the internal gradiometer signals during the first 400 ms from the stimulus onset, in the 3 cases (a, c, e) with identifiable phase reversal appearing at around 220 ms latencies (black arrows). Isofield maps at these latencies show clear dipolar field distributions (b, d, f). Isofield contours are stepped by 6, 8, 14 fT respectively. All cases were stimulated at 1000 Hz frequency.
F. Zappasodi et al. / Brain Research 917 (2001) 167 – 173
Fig. 4. AEP recordings (Cz derivation with bi-auricular reference, two repetition for each subject) of 2 preterm newborns, 38 weeks of gestational age. Traces are replicated twice to show the stability of neonatal peaks.
References [1] S. Azevedo, R.L. Longhini, Abdominal-lead fetal electrocardiographic R-waves enhancement, IEEE Trans. Biom. Eng. 27 (1980) 255–260. [2] A.J. Bell, T.J. Sejnowski, An information maximisation approach to blind separation and blind deconvolution, Neural Comput. 7 (1995) 1129–1159.
173
[3] S.M. Damstra-Wijmenga, The memory of the new-born baby, Midwives Chron. 104 (1991) 66–69. [4] K.J. Gerhardt, R.M. Abrams, Fetal hearing: characterization of the stimulus and response, Semin. Perinatol. 20 (1996) 11–20. [5] K.J. Gerhardt, X. Huang, K.E. Arrington, K. Meixner, R.M. Abrams, P.J. Antonelli, Fetal sheep in utero hear through bone conduction, Am. J. Otolaryngol. 17 (1996) 374–379. [6] P.G. Hepper, B.S. Shahidullah, Development of fetal hearing, Arch. Dis. Child. 71 (1994) F81–F87. [7] D.H. Hubel, T.N. Wiesel, The period of susceptibility to the physiological effects of unilateral eye closure in kittens, J. Physiol. 206 (1970) 419–436. [8] B. Johansson, E. Wedenberg, B. Westin, Fetal heart rate response to acoustic stimulation in relation to fetal development and hearing impairment, Acta Obstet. Gynecol. Scand. 71 (1992) 610–615. [9] N.P. Luz, Auditory evoked response of the human fetus: simplified methodology, J. Perinat. Med. 19 (1991) 177–183. [10] E.M. Ockleford, M.A. Vince, C. Layton, M.R. Reader, Responses of neonates to parents’ and others’ voices, Early Hum. Dev. 18 (1988) 27–36. [11] M.J. Parkes, P.J. Moore, D.R. Moore, N.M. Fisk, M.A. Hanson, Behavioral changes in fetal sheep caused by vibroacoustic stimulation: the effects of cochlear ablation, Am. J. Obstet. Gynecol. 164 (1991) 1336–1343. [12] D. Querleu, X. Renard, F. Versyp, L. Paris-Delrue, G. Crepin, Fetal hearing, Eur. J. Obstet. Gynecol. Reprod. Biol. 28 (1988) 191–212. [13] D. Querleu, R. Xavier, C. Boutteville, G. Crepin, Hearing by the human fetus?, Semin. Perinatol. 13 (1989) 409–420. [14] M. Samonas, M. Petrou, A.A. Ioannides, Identification and elimination of cardiac contribution in single-trial magnetoencephalographic signals, IEEE Trans. Biom. Eng. 44 (1997) 386–393. [15] S. Shahidullah, P.G. Hepper, Frequency discrimination by the fetus, Early Hum. Dev. 36 (1994) 13–26. [16] Z.D. Smith, L. Gray, E.W. Rubel, Afferent influences on brainstem auditory nuclei of the chicken: n. laminaris dendritic length following monaural conductive hearing loss, J. Comp. Neurol. 220 (1983) 199–205. [17] H. Sohmer, S. Freeman, Functional development of auditory sensitivity in the fetus and neonate, J. Basic Clin. Physiol. Pharmacol. 6 (1995) 95–108. ¨ [18] P. Strobach, K. Abraham-Fuchs, W. Harer, Event synchronous cancellation of the heart interference in biomedical signals, IEEE Trans. Biom. Eng. 41 (1994) 343–350. [19] M. Sur, A. Angelucci, J. Sharma, Rewiring cortex: the role of patterned activity in development and plasticity of neocortical circuits, J. Neurobiol. 41 (1999) 33–43. [20] R.T. Wakai, C. Arthur, M.S. Leuthold, B.M. Chester, Fetal auditory evoked responses detected by magnetoencephalography, Am. J. Obstet. Gynecol. 174 (1996) 1484–1486. [21] B. Widrow et al., Adaptive Noise Canceling: Principles and Applications, Proc. IEEE 63 (1975) 1692–1716. [22] R.O. Wong, The role of spatio-temporal firing patterns in neuronal development of sensory systems, Curr. Opin. Neurobiol. 3 (1993) 595–601. [23] V. Zarzoso, A.K. Nandi, E. Bacharakis, Maternal and foetal ECG separation using blind source separation methods, IMA J. Math. Appl. Med. Biol. 14 (1997) 207–225.