Although both EEG and MEG potentials are due to ionic current flow in neurons and their connections, the EEG is a record of voltage whereas the MEG is a record of current. In both cases, one is left with trying to solve what has been termed "the inverse problem," that is, trying to determine the source of these potentials based on signals reflected at the surface of the brain. In one of the special lectures, Dr. Samuel Williamson, of New York University (NYU) Medical School, made it clear that although early studies of the MEG had focused on temporal relations like EEG studies, multiple sensors now make it possible to use magnetic field maps "to yield quantitative information about the underlying source, without need for numerical analysis that takes account of the electrical properties of the intervening tissue." Furthermore, he stated that MEG localization procedures are revealing the sources of alpha activity in the brain and even complex tasks such as mental imagery can be monitored by measurements of the topographical distribution of the suppression of alpha rhythm.

It became clear early in the meeting that neuromagnetic topographic mapping and electroencephalographic topographic mapping are somewhat competing technologies. One of the main arguments presented by the neuromagnetic proponents was that MEG data have proven superior in the localization of the underlying neural sources in the brain. There were numerous oral and poster presentations directed at the problem of source localization; it became apparent that magnetic fields can be modeled using electromagnetic theorems and computer simulations can be used to study current dipoles. In a session on modeling and source localization of MEG activities, Dr. Shoogo Ueno showed the results of computer models of neural sources as single or multiple dipoles located in an inhomogeneous spherical conductor. His computer simulation results suggested four major points: (1) flux reversal phenomena can be observed in MEG topographic patterns in special cases where conductivities of the inhomogeneous regions change with pathological situations; (2) radically oriented

dipoles produce magnetic fields outside the inhomogeneous sphere; (3) magnetic fields on the sphere generated by single dipoles are sensitive to the inhomogeneities; and (4) a pair of opposing dipole or circular current sources produces only very weak electric fields but generates high magnetic fields, which are less affected by inhomogeneities in the head. Essentially, these data imply that changes in tissue conductivity such as could be caused by brain lesions or tumors can be detected by the MEG recordings and, in certain cases, the electrical signals could be inversely related to the magnetic field potentials.

There was also a series of oral presentations and posters related to mathematical models and computer simulations of EEG data. One of the presentations by Drs. Michel, Henggeler, and Lehmann of the University Hospital, Zurich, Switzerland, pointed out that EEG recordings are typically assessed in the frequency domain via FFTtransformation, but the resulting power maps are not directly suitable for dipole source localizations, since the phase information is lost. These investigators proposed that the amplitude and phase angle for each frequency point can be displayed in a sine-cosine diagram; this information can then be used to determine the dipole source localization.

MAJOR ADVANCES IN ANALYSIS TECHNIQUES

There are basically two major signals that represent neural transmission within the nervous system that can be recorded as surface potentials: (1) the continuous background activity, which typically involves potentials in the microvolt range and covers a bandwidth from dc up to 90 Hz, and (2) evoked potentials, which are discrete responses to external sensory or motor stimuli. Early studies of the EEG involved a limited number of channels of information and analysis of the data

into different frequency bands. However, within the last 10 years, the increase in computer technology has made multichannel recordings of 20 to 32 channels fairly common in EEG research. Commercially available systems contain software which provides for on-line analysis of the power spectrums and computes absolute power, relative power, coherence, amplitude, and phase. In addition, color graphic outputs allow the investigator to generate topographic maps of these data.

There were a number of papers that stressed the application of statistical procedures for understanding the underlying structure of the EEG. Dr. Harner, Medical College of Pennsylvania, presented a technique using a multivariate linear analysis technique called "single value decomposition." Essentially, this technique breaks the data down into a spatial matrix composed of a temporal vector, a spatial vector, and an amplitude vector. This approach is similar to that used in vector cardiography. Dr. Harner concluded that 90% of the variance was contained within these three wave forms. One of the most intriguing invited lectures titled "Beyond Topographic Mapping: Toward Functional-Analysis Imaging with 124 Channel EEG's and 3-D MRI's" was presented by Dr. Givens, director of the EEG Systems Laboratory, San Francisco, California. It is Dr. Givens' thesis that functional information from EEGs can be integrated with neuroanatomical information from magnetic resonance images (MRI), yielding information, with high temporal resolution, about the localization of function in the brain. Using 124 scalp electrodes, his laboratory has demonstrated equivalent dipole localization in the visual, auditory, and somatosensory modalities. The localization had an accuracy at least equal to that obtained with MEG. Furthermore, the EEG data collection and

analysis equipment cost about 25 times less ($100K) than a 37-channel MEG system. Figure 4 shows equivalent dipoles (small dark cylinders) for elecdipoles (small dark cylinders) for electrical stimulation of left and right middle fingers and the index finger. The dipoles have been registered with a model of the subject's brain shown in outline inside the head. The outer 11 millimeters of the model have been stripped away to visualize the three-dimensional position of the dipoles. The face and head surfaces were also reconstructed from the subject's MRI brain scan. It is evident that the dipoles are correctly positioned. Although single dipoles are convenient for characterizing simple sensory stimuli, they are not suitable for describing higher cognitive functions since it is well-established that even simple cognitive processes are widely distributed in the brain. EEG Systems Laboratory has developed a method called event-related covariance analysis (ERC) to characterize the parallel, distributed activity of cognition. The method is based upon an empirically tested mathematical model of brain neural action developed by Professor Walter Freeman at U.C. Berkeley which concludes that the macro-potentials of functionally related cortical regions will be highly coherent. The theory and method have been validated in many published experiments in animals and humans. The lines on the head shown in Figure 5 represent all of the ERCS, which are computed in each fraction-of-a-second analysis. Using this technique, Dr. Givens has shown that distributed neuronal networks are affected by sustained mental work several hours before performance deteriorates significantly. In a study with pilots performing a difficult task, of long duration, striking changes occurred in the ERC patterns changes occurred in the ERC patterns after the pilots performed the task for 7 to 9 hours (middle) but before performance deteriorated (see Figure 6).

Using neural-network pattern recognition technology, it was possible to distinguish the state of incipient performance impairment with an accuracy of over 80%. Dr. Givens believes that further development of the technology could lead to an early warning system to detect leading indicators of impaired performance in aircraft pilots or operators of other complex equipment.

APPLICATIONS OF TOPOGRAPHIC MAPPING

As discussed by Dr. Lehmann, Department of Neurology, University Hospital, Zurich, Switzerland, “mapping has proven to be a superior way to comprehensively display threedimensional data, e.g., multichannel EEG and MEG data, as momentary maps, FFT band power maps, momentary event-related evoked potential maps, and results of statistical electrodeby-electrode map comparisons."

In the case of normal human subjects, EEG and MEG data are being assessed as methods for examining individual differences in brain function, which correlate with one's ability to process information and solve problems. Furthermore, the effects of stress, drugs, environmental factors, etc. can be evaluated using topographic techniques. A number of papers dealing with topographic analysis and evoked potential techniques discussed significant changes correlated with learning and cognitive function. An interesting finding by Dr. Yamamoto of Dokkyo University, Japan, and his colleagues was that EEG studies could be used to determine the duration of work and rest periods. In their studies, operators of visual display terminals were subjected to psychological stress; they found that when the subjects complained of fatigue there was an increase in longlasting theta waves in the EEG. Following a rest period the EEG returned to normal.

MAGNETIC STIMULATION OF THE CENTRAL NERVOUS SYSTEM

Transcranial pulsed magnetic stimulation of the brain was first described by Barker et al. in 1985 (Ref 5). Single coils were used for these studies and, as a result, broad areas of the brain were stimulated simultaneously. At about the same time, Ueno et al. (Ref 6) developed a system composed of a pair of coils that could concentrate the induced eddy currents in a small area. Essentially, a pair of coils is positioned outside the head such that the timevarying magnetic fields pass through the head in opposite directions around

the target to be stimulated (see Figure 7). A pair of coils with windings of five turns of 5-mm-wide by 2-mm-thick copper wire is excited by discharge currents (4,000 to 7,500 A) that pass through a capacitor bank (600 to 900 V). The discharged currents produce timevarying magnetic fields which give rise to eddy currents of 50 A/m2 with a duration of 0.1 ms. These currents are concentrated at a point 10 mm below the cross point of the paired coil; in the case of brain stimulation the resolution is within 5 mm. There were a number of papers presented at the meeting (using this technique) that reported the effects of magnetic stimulation in both animals and humans.

Computer simulations and models of neural excitation have been developed which suggest that it will be possible to functionally and anatomically map the motor systems of the brain. Following the meeting, the senior author of this article (J.W. Wolfe) visited Dr. Ueno's laboratory at Kyushu University and experienced the magnetic stimulation of peripheral nerves in the arm and hand (I declined the offer of brain stimulation). This was a dramatic experience since, unlike direct electrode stimulation with needles placed in the nerves (which is the conventional method presently used in neurological evaluations), the stimulation was completely painless and without any sensation. In