Patent No. 5230344 Evoked potential processing system with spectral averaging, adaptive averaging, two dimensional filters, electrode configuration and method thereof

 

Patent No. 5230344

Evoked potential processing system with spectral averaging, adaptive averaging, two dimensional filters, electrode configuration and method thereof (Ozdamar, et al., Jul 27, 1993)

Abstract

The evoked potential processing system includes, in one embodiment, a spectral averaging method. Time based, digital pre-stimulus and post-stimulus electroencephalographic (EEG) signal streams are obtained and are converted into frequency spectrum signals. A differential spectrum is obtained. The differential spectrums from a plurality of sweeps are summed. The summed differential spectrum is then converted into a time based signal stream which contains the evoked potential (EP) signal therein. The EP signal can also be obtained utilizing a two-dimensional filter. Pre- and post-stimulus EEG signal streams for a sub-group of stimuli, wherein each stimulus in a group has the same intensity or frequency, are filtered by conventional averaging or spectral differential averaging. The time based, filtered, post-stimulus EEG signal streams are placed in an array and the array is then filtered by a two-dimensional Fast Fourier Transform (FFT) filter. The array is then filtered by a mask and the masked array is then transformed into a time based format by an inverse FFT. The adaptive averaging technique utilizes a computational formula which computes an estimated running signal to noise ratio. When the difference between the pre-stimulus running SNR and the post-stimulus running SNR is less than a predetermined threshold, further stimulation and acquisition of EEG signals stops. Hence, the post-acquisition processing of the EEG signals is limited to that number of EEG sweeps. The electrode wire configuration uses a cross wiring scheme wherein the shield of a particular wire is connected to the other electrode wire to eliminate artifacts in the respective electrode wire.

Notes :

Evoked potential processing system with spectral averaging, adaptive averaging, two-dimensional filters, electrode configuration and method thereof. Filed July 1992, granted July 1993. Shows ways to collect, identify and store in memory various evoked potentials. Done through use of real time correlation of pre-determined stimuli and evoked potential responses. This is a cause - effect type scenario. Pre-determined could be anything from radio and television programs to 'street theater'. This could also be done with remote neural monitoring and storage of what was monitored then correlating this with say a video camera from work, all that would be needed is an exact time frame of a known occurrence within the video to go from, in either direction, with what was earlier recorded. Can also collect, identify and store evoked potentials from more than one area of the brain simultaneously. Also goes into varying frequencies of stimuli, i.e. brightness or volume. This would/ could account for 'whisper' type attacks and vary degrees of clarity and shading in the image attacks. Take into account a much better computer being used.

BACKGROUND OF THE INVENTION

The present invention relates to an evoked potential processing system. The processing system includes special processing methods such as spectral averaging, variable sweep averaging or adaptive averaging, two-dimensional filtering, and/or a special electrode wire configuration.

The brain wave activity of a patient or a living being may be monitored by sensing electroencephalographic (EEG) signals sensed on the skull of the patient or living being. Essentially, electrodes are placed on the skull in order to sense EEG activity. The response of a brain to stimulus is indicative of the condition of the subject. For example, auditory brain stem response can be used as a tool to assess the auditory function in certain patients, particularly infants. EP signals can also be used to detect the subject's condition during surgery. Auditory brain stem response (ABR) reveals information regarding the peripheral organ, that is, the ear, and the brain stem auditory pathways. Since ABR is non-invasive and does not require anesthesia or sedation, such tests can be easily administered in office environments or in new born nurseries. In some instances, ABR measurements are obtained utilizing auditory clicks wherein each click has a certain pulse width, a certain intensity or amplitude level and covers a broad frequency band. Alternatively, an ABR measurement can utilize frequency specific audio stimuli such as tone bursts for special testing paradigms or test patterns.

Rather than use audio stimuli to affect EEG activity, visual stimuli could be used. Flashes of light (broad band frequency spectrum), intensity specific visual stimuli, hue or color stimuli, frequency specific stimuli and checkerboard pattern visual stimuli could be used. Further, somato sensory stimuli could be utilized, such as tactile or physical stimulus or electrical pulse stimulus. These types of stimuli affect EEG activity and such stimuli can be used to test brain stem pathways and peripheral sensory activity.

Auditory brain stem response (ABR) tests using click evoked ABR thresholds have been quite effective in determining average hearing sensitivity in the mid frequency (2-4 kHz) range. However, these click evoked ABRs give little information about the shape of an audiogram. Latency, the time delay of certain waveform shapes in the EEG signal stream obtained during an ABR test, are considered better estimates of the audiogram. These estimates of the audiogram can be used to generate a family of audiograms since there can be more than one audiogram producing the same latency-intensity functions.

Rapid acquisition of evoked potentials can be achieved primarily by three approaches: a) employing signal enhancement processing methods, b) using high stimulus rates, and c) testing both ears simultaneously. The first approach requires an innovative process other than the traditional averaging. Traditional averaging normally involves stimulating EEG activity with multiple, timed stimuli, obtaining the time based EEG signal streams with a time base reference link to the application of the stimulus, adding and averaging coincidental time segment EEG signals, both pre-stimulus and post-stimulus EEG signals. Limitations of the traditional ensemble averaging method in evoked potential acquisition have long been recognized. The problem arises primarily from the poor signal to noise ration (SNR), the nonstationary nature of the noise (that is, the noise may move with time) and the small amplitude of the signal response.

The second approach to speed up the EP acquisition process requires the increase of the stimulus repetition rate. This, however, generally changes the physiological characteristics of the system reducing the usefulness of the responses. Typically, at high stimulus rates neural adaptation takes place and responses are reduced and prolonged. This effect, however, can be reduced or eliminated by taking advantage of the special characteristics of the sensory organ. For the auditory system, intensity and frequency of tonal stimuli can be ordered such that adaptation is kept at a minimum.

A third approach is to use techniques that will allow the simultaneous recording from both ears thus effectively halving the testing time for each ear. Recordings from concurrent presentation of slightly different rate sound stimuli to both ears can be averaged and evoked potentials, EPs, can be separately obtained.

Automated response detection techniques can be broadly classified in two groups: EP signal statistic-based methods and EP waveform-based methods. Statistic-based methods compute various statistical measures of individual and average responses across time and sequences to detect the presence of a response. F.sub.sp is a statistical approach using variance analysis in calculating the ratio of the ABR to the estimated averaged background noise. See the article entitled "Objective Detection of Averaged Auditory Brainstem Responses" by M. Don, et al. and the article entitled "Quality Estimation of Average Auditory Brainstem Responses" by C. Elberling, et. al. Waveform-based methods detect the presence of a response by comparing the test waveform to another waveform either learned previously by the system or acquired under similar or no-stimulus conditions.

While the response recognition problem has been given adequate attention, threshold determination procedures have not received similar attention. This may be due to the fact that under laboratory conditions in which response artifacts are minimal, threshold determination reduces to the trivial problem of level detection.

With current prior art devices, the recording time necessary to determine the ABR threshold for a given stimulus is about 20 minutes (10 recordings, 1024 sweeps). If a four-frequency audiogram is desired, about 80 minutes of recording time are required. For two ears the testing time approaches three hours.

Most processing techniques are configured to obtain a predetermined number of post-stimulus EEG or raw EP signal during the averaging routine. The predetermined number of EEG signals are summed such that substantially identical time based signal segments, referenced to the application of the stimulus for that particular signal sweep, are added together. In order to improve the signal to noise ratio, it is generally thought that the number of signals averaged, n, should be increased. It has been proposed that an SNR value can be estimated using different formulas proposed by various researchers (Mocks et al., 1984; Turetsky et al., 1988). Also, the use of unbiased estimators has been proposed by Mocks et al.

Where:

SNR=Running signal to noise ratio estimate

P.sub.s =Average signal power

P.sub.n =Average noise power

X.sub.k (t)=k.sup.the EEG sweep

K=number of sweeps averaged

T=number of data points in each sweep. ##EQU1##

As the equations are formulated, it is extremely difficult to compute SNR in real-time with a running average. In addition to requiring many computations, all the single responses must be kept in memory. The computations take longer and longer as the number of averaged sweeps increases. The prior art references did not take into account both pre-stimulus and post-stimulus EEG signals but dealt with only post-stimulus EEG signals.

In addition to being averaged by traditional summation techniques, the EEG signal streams may be filtered with a two-dimensional filter. Two-dimensional filtering is based on the idea that image processing methods can reveal prominent events in an array of consecutive EPs and suppress transient artifacts which occur in individual recordings only. Two-dimensional filtering follows the general principles of image processing which are well known and commonly applied to pictorial type problems (Gonzales & Wintz, 1987). However, evoked potential or EP signals are not normally recognized as including identifiable images or pictures. Accordingly, such multi-dimensional signal processing techniques are seldom used in processing bioelectric signals. Sgro et al. (1985) first proposed two-dimensional filtering for EP reconstruction to track dynamic changes. Sgro's paper discloses the use of a two-dimensional Fast Fourier Transform (FFT) filter to process multiple, EP or EEG signal streams. However, the stimuli used to develop the EP signals were uniform, that is, the intensity and frequency (either click or tone burst) of each stimulus was substantially identical. Sgro first transformed the EP signals using a two-dimensional FFT, then used a mask to drop certain frequencies from the transformed two-dimensional array (those frequencies which exceeded certain pre-determined limits) and then retransformed the array with an inverse FFT to obtain filtered, EP signal streams in the time domain.

OBJECTS OF THE INVENTION

It is an object of the present invention to provide a system and a method for extracting evoked potential signals from a stream of pre- and post-stimulus EEG signals.

It is another object of the present invention to provide a system which generates certain information such as one or more audiograms or one or more latency-intensity curves during an auditory brain stem response in about 5-10 minutes per ear.

It is a further object of the present invention to remove artifacts from EP-EEG signal stream in real time.

It is another object of the present invention to utilize a variable sweep averaging or adaptive averaging using a computational formula to estimate the running signal to noise ratio requiring the storage of the running sum of the time base EEG signal and the square of that running sum.

It is another object of the present invention to utilize a spectral enhancement averaging technique to remove or reduce noise in the EP signals.

It is an additional object of the present invention to utilize a two-dimensional filter to remove noise from the EEG signal streams particularly when the chain of stimuli has increasing levels of intensity and/or frequency.

It is a further object of the present invention to remove artifacts and noise from the EEG evoked potential signals by combining the spectral enhancement averaging and the two-dimensional filter and, hence, identify the evoked potential (EP) signals in the EEG signal streams.

It is an additional object of the present invention to provide an electrode wiring configuration which reduces or eliminates artifacts in the EEG signal carried by the electrode wiring.

SUMMARY OF THE INVENTION

The evoked potential processing system includes, in one embodiment, a spectral averaging method. Time based, digital pre-stimulus and post-stimulus electroencephalographic (EEG) signal streams are obtained and are converted into pre- and post-stimulus frequency spectrum signal streams. A differential spectrum is obtained from each pre- and post-stimulus signal. The differential spectrums from a plurality of sweeps are summed by correlating respective time bases in each differential spectrum and summing coincidental time segments together. The summed differential spectrum is then converted into a time based signal stream which contains the evoked potential (EP) signal therein. The raw EP signal can also be processed utilizing a two-dimensional filter. Pre-and post-stimulus EEG signal streams for a sub-group of stimuli, wherein each stimulus in a group has the same intensity or frequency, are filtered by conventional averaging or spectral differential averaging. The time based, filtered, post-stimulus EEG signal streams are placed in an array and the array is then filtered by a two-dimensional Fast Fourier Transform (FFT) filter. The masked array is then transformed into a time based format by an inverse FFT. The adaptive averaging technique utilizes a computational formula which computes an estimated running signal to noise ratio for the pre- and post-stimulus EEG signal by storing the running sum of a time base segment of each signal stream as well as the square of the time base segment. When the difference between the pre-stimulus running SNR and the post-stimulus running SNR is less than a predetermined threshold, further stimulation and acquisition of EEG signals stops. Hence, the post-acquisition processing of the EEG signals is limited to that number of EEG sweeps. The electrode wire configuration uses a cross wiring scheme wherein the shield of a particular wire is connected to the other electrode wire to eliminate artifacts in the respective electrode wire.

--------------------------------------------------------------------

The claims appended hereto are meant to cover modifications and changes within the spirit and scope of the present invention.