Source Analysis of auditory evoked
electromagnetic fields

Terence W. Picton
Rotman Research Institute, University of Toronto
3560 Bathurst St., Toronto, Ontario, CANADA M6A 2E1

Abstract: The responses of the human auditory cortex may be recorded using electric or magnetic sensors at the scalp.  The largest responses are evoked by changes in the auditory input.  Detecting these changes is essential for monitoring objects in the auditory world.  Different types of change detection are shown in the N1 wave to the onset of a stimulus, the mismatch negativity to a change in a repeating stimulus, and the object-related negativity evoked by the occurrence of a new auditory object.  Rhythmic responses of the auditory cortex reflect interactions between the thalamus and cortex.  These rhythmic responses may relate to the binding together of auditory information into a conscious perception. 

ELECTRIC AND MAGNETIC RECORDINGS

The activity of neurons within the brain can be recorded by measuring the electromagnetic fields from the scalp [1].  These fields are generated by the flow of current within activated neurons. The fields vary with the number of neurons activated, the synchronicity of their activation and the geometric congruity of the resultant current vectors.  The alignment of the pyramidal cells within the cerebral cortex makes them particularly able to generate electromagnetic fields at a distance.  When an excitatory synapse is activated in the dendrite of such a neuron, current flows into the neuron (Figure 1).  This current then diffuses through the intra-cellular spaces and returns to the extracellular space at a distance from the synapse.  This flow of current sets up sources and sinks in the extracellular space. Since the extracellular currents associated with the sources and sinks are oriented in opposing directions, the magnetic fields that they create cancel each other out.  However, the intracellular current between the region of synaptic activation and the region where the currents return to the extracellular space remains uncancelled, and generates magnetic fields that can be recorded at a distance [2].  The magnetic fields circle around the current according a right hand rule – in the direction of the fingers when the current is grasped with the thumb pointing in the direction of the current.

The magnetic fields generated by intracelluar currents can be recorded using super-conducting quantum interference devices as the magnetoencephalogram (MEG). If the intracellular currents are oriented radially to the surface, as might occur when the activated neurons are on a cortical gyrus, the fields circling around the direction of the current flow are equal at all locations on the surface of the roughly spherical head.  If, on the other hand, the intracellular currents are oriented tangentially, as might occur if the activated neurons are within a sulcus, the magnetic field exits the head on one side of the activity and enters on the other side.

The extracellular currents associated with neuronal activation set up potential fields that can be recorded from the scalp using simple electrodes as the electroencephalogram (EEG) [3].  As shown in Figure 1, the electric recordings from a tangential current dipole are oriented orthogonally to the magnetic field.  A major advantage of MEG is that the magnetic fields are not affected by the volume conductor.  For example, changes in the thickness of the skull over different regions of the cortex do not affect the measurements.  In addition, the component structure of the MEG response is usually simpler than the EEG response since the MEG is blind to radial currents.  Both these factors make determining intracerebral sources of scalp-recorded evoked responses more accurate for MEG than for EEG. MEG is easier to record than the EEG since the sensors are simply placed adjacent to the scalp and do not have to be individually applied. The disadvantages of the MEG compared to EEG are the cost and the inability of MEG to detect radial or deep current sources.  Since the MEG and EEG complement each other, their combined use can greatly improve our understanding of what is occurring in the brain when sounds are processed [4].

Figure 1.Generation of electric and magnetic fields.

AUDITORY EVOKED ELECTROMAGNETIC FIELDS

The many different electromagnetic responses evoked from the human brain by auditory stimuli can be classified as transient, sustained or steady state [5].  Transient responses are brief changes in the electromagnetic fields evoked by the onset or offset of a stimulus or by some change in an ongoing stimulus.  These responses are usually recorded in the time-domain as a waveform with peaks and troughs occurring at latencies from one to several hundred milliseconds. Sustained potentials are continuous shifts in the recorded level that last through the duration of the stimulus. Steady state responses are responses to recurring stimuli that have settled so that they maintain a stable amplitude and phase relationship to the stimuli.  These responses are often recorded in the frequency domain as peaks in a spectrum, with frequencies that are harmonically related to the stimulus rate. 

Transient responses are characterized by a polarity and either a location in a sequence of waves or a characteristic latency. For example, the N1 or N100, the first prominent negative wave in the slow onset-response recorded from the vertex, has a characteristic peak latency of 100 ms [6]. The magnetic counterpart of this wave, called the N1m or N100m, is recorded as an outgoing field on one side of the current dipole and an ingoing  field on the other side.  Which side is which depends on the hemisphere wherein the currents are located.  As shown diagrammatically in Figure 2, the N1m generated in the right auditory cortex is recorded as an ingoing field posteriorly and an outgoing field anteriorly. The main generators for these late transient responses are in the cortex on the top of the temporal lobe. Sustained potentials (SP) or sustained magnetic fields (SF) are also generated in this area, usually with a source location more anterior than that of the N1 wave [7].  

Figure 2. Transient and sustained auditory responses.

Auditory steady state responses can be recorded in response to stimuli at many different presentation rates [8]. The most easily recorded auditory steady state response is the 40 Hz response, which has been used to assess hearing and to evaluate the level of anesthesia.  The rhythmic nature of the steady-state responses is related to the spontaneous activity of the cortex.  Gamma rhythms in the brain (with frequencies of 25-50 Hz) have been associated with perceptual processes such as the binding together of different attributes into a perceived object [9].  These gamma-band rhythms may be “spontaneous”, “induced” by a stimulus as a burst of activity that is not specifically phase locked to the stimulus, “evoked” in a time-locked manner by a stimulus, or “emitted” in association with some perceptual or motor decision [10].  In addition they may be “driven” by stimuli that are presented at rates within the gamma-frequency band. 

SOURCE ANALYSIS

The activation of neural tissue causes fluctuations in both the cerebral blood flow and the electromagnetic fields.  These changes can be recorded from sensors placed on or near the scalp. Although hemodynamic recordings map the magnitude of blood flow in each region of the brain, electrical and magnetic recordings often map the raw output of the sensors – the waveforms recorded at different locations on the scalp. Source analysis of these scalp recordings should be able to delineate the intracerebral generators of these electric or magnetic fields.  However, more than one intracerebral current distribution can produce the same field topography at the scalp: the electromagnetic inverse solution is non-unique and constraints are necessary for the analysis.  Two models are used for the inverse solutions: discrete and distributed. 

Discrete models search for a solution with relatively few point current-dipoles each representing an active area of the brain.  Brain Electromagnetic Source Analysis (BESA) [11] uses dipoles with fixed locations and orientations that vary in strength across time.  The locations and orientations of the dipoles can be fixed a priori on the basis of the known anatomy and connections of the brain, can be ‘seeded’ from the locations known to be active from functional magnetic resonance imaging (fMRI), or can be determined using iterative fitting protocols. Distributed models search for a general current distribution within the brain volume.  The Minimum Norm solution [1] is constrained to have the minimum total current and Low Resolution Electromagnetic Tomography (LORETA) [12] is constrained to have the spatially smoothest current distribution.  Since these constraints are imposed mathematically, the solutions require little or no human input.  However, the constraints may not always represent what is happening in the brain.  The currents in adjacent regions of the cortex may be quite distinct from each other and may not fit with constraints such as minimum overall current or maximum spatial smoothness. 

RESPONSES TO CHANGE

The transient auditory responses of the human cortex process the changes that occur in our auditory world.  The stimuli used to evoke the responses analyzed in Figure 3, were occasional (1/5 seconds) onsets of a brief tone, offsets of a continuous tone or brief changes in the frequency of this tone.  All evoke a complex of waves that is characterized electrically by a negative wave (N1) recorded maximally over the frontocentral scalp with a peak latency of just over 100 ms. This wave occurs in response to the offset (open triangle) as well as the onset (filled triangle) of a stimulus, although for the same stimulus duty-cycles, the offset response is much smaller than the onset response.  A source analysis of this wave shows that it derives mainly from activity in the supratemporal plane of both hemispheres (sources 1 and 2) with the left side (thin line) responding earlier to the right-sided stimuli. The locations of these sources are likely a little posterior to the primary auditory cortices on Heschl’s gyri.  The sources likely reflect activity in auditory association areas. However, the analysis is based on equivalent dipoles and these may represent the simultaneous activation of multiple auditory areas on the supratemporal plane. In addition to the vertically oriented dipoles there are radially oriented dipoles (sources 3 and 4) that show activation at a slightly later latency.  These likely reflect the activation of the auditory cortices on the lateral wall of the temporal lobe. Further analyses may show subsequent activation of regions of the frontal lobe [13], although the extent of these activations will be determined by the amount of attention being paid to the stimuli. The pattern of activation in these frontal areas overlaps with the activation of the temporal lobe, and it is difficult to disentangle the different generators unless recordings with high signal-to-noise and multiple electrode locations are used.

Figure 3. Source analysis of auditory N1 wave

If a mistuned harmonic occurs in a complex sound, this is heard as a separate auditory object since most objects in the real world generate sounds with harmonically related components. Evaluating the response to this type of stimulus may thus give us some idea of how the brain performs a basic auditory scene analysis.  The mistuned stimulus evokes an N1 wave and a small additional  “object related negativity” that also has its main source on the supratemporal plane [14].

If some feature of a repeating standard auditory stimulus occasionally changes, the auditory response to the deviant stimulus contains, in addition to the usual N1 wave, a small “mismatch negativity” [15].  Noticing this stimulus requires not only the ability to discriminate the difference but also a memory of the previous stimulus. The source of this wave is mainly on the supratemporal plane but it is often a little anterior to the source the N1 wave when the deviance is a change in the tonal frequency of the stimulus (Figure 4) [7].  The mismatch negativity is often recorded using rapid stimulus rates.  In Figure 4, the stimulus rates were 4/second. At these rates the N1 responses to the standard stimuli (asterisks) are in a refractory state and are so small as to be barely recognizable.  The response to the standard stimulus following the deviant stimulus (open triangle) shows a small enhancement of its N1 since the refractory state from the rapid stimulation is frequency-specific and this standard stimulus differs from the preceding stimulus.  The response to the deviant stimulus shows a negative wave with a latency longer than the normal N1. Subtracting the response to the standard stimulus from the response to the deviant stimulus gives the mismatch negativity (filled triangle).  The mismatch negativity is very sensitive to irregularities is recurring stimuli and can pick out changes in the patterns of stimulus repetition as well as changes in simple characteristics. 

Figure 4.  Source analysis of mismatch negativity

RHYTHMIC RESPONSES

The auditory steady state responses at rates near 40 Hz are generated by interactions between the thalamus and cortex [8]. MEG studies have shown that the build up of this steady state response occurs with a similar time-course to the integration of stimulus information into perception [16].  Figure 5 shows a source analysis of electrically recorded responses.  The electric fields recorded from the scalp represent the superimposed activity of cortical and thalamic sources.  The main generators are in the thalamus (source 1) and supratemporal plane (sources 3 and 4). Sources 3 and 4 occur about 15 ms after source 1.  The later activation of the cortical response would fit with the usual activation pattern.  However, it is also possible that feedback oscillations occur between the thalamus and cortex. In addition, there are generators on the lateral wall of the temporal lobe (sources 5 and 6) that are activated about 5 ms later than source 1.

Figure 5.  Source analysis of 40 Hz response.

CONCLUDING COMMENTS

The studies reviewed in this paper show how the human auditory cortices in the temporal lobe detect changes in the auditory world.  Studies of the steady state responses show that cortical activity occurs in combination with the thalamus.  These measurements show what happens when sounds are heard passively and processing occurs automatically.  When attention is paid to the auditory input, the response waveforms become more complex and many more sources in the brain become active. 

Electromagnetic recordings can show the time course of activation in the human auditory system.  However, MEG and EEG lack the anatomical precision of other functional neuroimaging techniques that measure local cerebral blood flow, such as fMRI.  It is often difficult to determine the exact sources of the scalp-recorded electromagnetic fields.  This is particularly true when multiple regions of the brain are simultaneously active.  We can usually determine the main sources of activity in the auditory cortices of the temporal lobe but later activations are less distinct.  

Multiple areas of activation can be seen quite clearly in fMRI recordings. However, hemodynamic changes occur between several hundred milliseconds and several seconds after the neuronal activity.  Until recently, most studies of cerebral activation looked at periods lasting several to many seconds. Over the past few years, it has become possible to make hemodynamic measurements for specific trials using the same selective averaging procedures as for the electromagnetic recordings, but these new measurements are still limited by the several-second time-constant of the blood flow changes and the long inter-stimulus intervals required for the blood flow measurements to return to baseline.

To understand the processes of perception and cognition, we need to determine exactly when a particular region of the brain is active.  Is an active area continuously active or just active at one period of time?  If active at one period of time, is this before or after another active area?  Electric and magnetic recordings can provide the timing. Hemodynamic recordings can provide the localizations and these may help constrain the electromagnetic source analyses.  The future is bright for the combined use of EEG, MEG and fMRI to understand how the human brain hears sounds, identifies auditory objects and understands what they mean. 

Acknowledgments: This research program was supported by the Canadian Institutes of Health Research. 

REFERENCES

[1] M. Hämäläinen, R. Hari, R. J. Ilmoniemi, et al., “Magnetoencephalography - theory, instrumentation, and applications to noninvasive studies of the working human brain,” Reviews of Modern Physics, vol 65, pp. 413-497, 1993.

[2]  S. J. Williamson, and  L. Kaufman,  “Analysis of neuromagnetic signals,” in A. S. Gevins and A. Rémond (Eds.), Handbook of Electroencephalography and Clinical Neurophysiology:  Volume 1. Method of Analysis of Brain Electrical and Magnetic Signals (pp. 405-448). New York: Elsevier, 1987

[3] T. W. Picton, O. Lins, and M. Scherg, “The recording and analysis of event-related potentials,” in F. Boller and J. Grafman (Eds), R. Johnson Jr. (Section Ed.) Handbook of Neuropsychology. Volume 10. Section 14 Event-Related Brain Potentials and Cognition, (pp. 3-73), Amsterdam: Elsevier, 1995

[4] C. Pantev, O, Bertrand, C. Eulitz, et al. “Specific tonotopic organizations of different areas of the human auditory cortex revealed by simultaneous magnetic and electric recordings,” Electroencephalography and Clinical Neurophysiology, vol. 94, pp. 26-40, 1995

[5] T.W. Picton, “Auditory evoked potentials,” in D. D. Daly and T. A. Pedley (Eds.)  Current Practice of Clinical Electroencephalography, 2nd Edition (pp. 625-678) New York: Raven Press, 1990

[6] R. Näätänen, and T. W. Picton, “The N1 wave of the human electric and magnetic response to sound:  a review and an analysis of the component structure,”  Psychophysiology, vol. 24, pp. 375-425, 1987

[7] M. Scherg, J. Vajsar, and T. W. Picton, “A source analysis of the late human auditory evoked potentials,”  Journal of Cognitive Neuroscience, vol. 1, pp. 336-355, 1989.

[8] T. W. Picton, M. S. John, A. Dimitrijevic, et al “Human auditory steady-state responses,” International Journal of Audiology, submitted, 2002

[9] W. Singer, “Response synchronization: A universal coding strategy for the definition of relations,” In M. S. Gazzaniga (Ed.), The New Cognitive Neurosciences, (pp. 325-338). Cambridge: MIT Press, 2000

[10] R. Galambos,  “A comparison of certain gamma band (40-Hz) brain rhythms in cat and man.” in E. Basar and T. H. Bullock (Eds.), Induced Rhythms in the Brain (pp. 201-216). Boston: Birkhäuser, 1992

[11] M. Scherg, “Fundamentals of dipole source potential analysis,” in F. Grandori, M. Hoke, and G. L. Romani (Eds):  Advances in Audiology 6. Auditory Evoked Magnetic Fields and Electric Potentials. (pp. 40-69) Basel: Karger, 1990.

[12] R. D. Pascual-Marqui, C. M. Michel, and D. Lehmann, “Low resolution electromagnetic tomography:  A new method for localizing electrical activity in the brain.”  International Journal of Psychophysiology, vol. 18, pp. 49-65, 1994.

[13] T. W. Picton, C. Alain, D. L Woods, et al. “Intracerebral sources of human auditory evoked potentials,” Audiology and Neuro-Otology, vol. 4, pp. 64-79, 1999

[14] C. Alain, S. R. Arnott, and T. W. Picton, “Bottom-up and top-down influences on auditory scene analysis: Evidence from event-related brain potentials,” Journal of Experimental Psychology, Human Perception and Performance, vol. 27, pp. 1072-1089, 2001

[15] Picton, T. W., Alain, C., Otten, L., et al.  “Mismatch negativity:  Different water in the same river,” Audiology and Neuro-Otology, vol. 5, pp. 111-139, 2000

[16] Roß, B., Picton, T. W., and Pantev,C. The onset of the steady-state response reflects temporal integration in the auditory system. Hearing Research, in press.