Surgery – Diagnostic testing – Detecting brain electric signal
Reexamination Certificate
2001-10-26
2004-04-06
Nasser, Robert L. (Department: 3736)
Surgery
Diagnostic testing
Detecting brain electric signal
C600S300000, C600S546000, C600S547000
Reexamination Certificate
active
06718199
ABSTRACT:
FIELD OF INVENTION
The invention relates to the measurement of electrophysiologic responses, and more particularly to enhancing the signal-to-noise ratio in such measurements.
BACKGROUND
In making a diagnosis, it is often useful to have the patient's cooperation. This is particularly true in the diagnosis of disease involving sensory pathways to the brain. For example, a straightforward way to assess a patient's hearing is to simply ask the patient whether he can hear particular tones having various frequencies and amplitudes.
In many cases, one takes for granted that a patient will be able to answer such questions. However, in some cases, a patient cannot communicate his perception. This occurs most frequently when the patient is an infant, or when the patient is unconscious. In a veterinary setting, it is rare to encounter a patient that can accurately communicate perception at all.
One approach to evaluating an infant's hearing is to make a sound and to then measure an evoked response associated with that sound. This evoked response is typically an electrophysiologic signal generated in response to the sound and traveling between the inner ear and the brain along various neural pathways, one of which includes the auditory brainstem. This signal is thus referred to as the “auditory brainstem-response,” hereafter referred to as the “ABR.”
The ABR is typically only a small component of any measured electrophysiologic signal. In most cases, a noise component arising from other, predominantly myogenic, activity within the patient dwarfs the ABR. The amplitude of the ABR typically ranges from approximately 1 microvolt, for easily audible sounds, to as low as 20 nanovolts, for sounds at the threshold of normal hearing. The noise amplitude present in a measured electrophysiologic signal, however, is typically much larger. Typical noise levels range from between 2 microvolts to as much as 2 millivolts. The resulting signal-to-noise ratio is thus between −6 dB and −100 dB
One approach to increasing the signal-to-noise ratio is to exploit differences between the additive properties of the ABR and that of the background noise. This typically includes applying a repetitive auditory stimulus (a series of clicks, for example) and sampling the electrophysiologic signal following each such stimulus. The resulting samples are then averaged. The ABR component of the samples add linearly, whereas the background electrophysiologic noise, being essentially random, does not. As a result, the effect of noise tends to diminish with the number of samples. The number of samples required to reach a specified signal-to-noise level depends on the noise level present in the samples. In principle, therefore, one can achieve a specified signal-to-noise ratio either with a small number of relatively quiet samples or with a large number of relatively noisy samples.
In practice, signal averaging techniques such as that described above are unlikely to work when the signal-to-noise ratio is worse than −48 dB. Since a minimally acceptable 5% confidence level requires a signal-to-noise ratio of at least −4 dB, this signal-averaging approach is prone to inaccuracy.
Signal averaging methods as described above perform best when the background noise is relatively constant. For example, the steady drone of an air-conditioner can readily be separated from a signal of interest. Such background noise is referred to as “stationary” noise.
The noise component of an electrophysiologic signal is often non-stationary. For example, after a few minutes of taking measurements, an infant may begin to stir, thereby momentarily increasing the background electrophysiologic noise level. The infant might then return to a deep sleep, thereby reducing the background electrophysiologic noise level.
The non-stationary nature of the noise component poses a dilemma for a clinician attempting to measure the ABR. For example, if the infant begins to stir, the clinician might suspend taking measurements to avoid contaminating data already collected with noisy data. This might prove to be a good decision if the infant were to fall back into a deep sleep, since one could then acquire additional quiet samples. However, even noisy samples can improve signal-to-noise ratio, provided that there are enough of them available. Hence, this might also prove to be a poor decision if the infant were to continue stirring. In such a case, it would have been better to have acquired the additional, albeit noisy samples. Because the behavior of an infant is, to a great extent, unpredictable, the clinician occasionally makes an incorrect guess, thereby either wasting time or needlessly corrupting acquired data.
SUMMARY
The invention is based on the recognition that, by dividing the sequence of samples that make up the signal into subsequences of samples, one can reduce the signal-to-noise ratio of an electrophysiologic signal and avoid many difficulties posed by the presence of non-stationary noise. The samples within a particular subsequence are characterized by a common range of values of a sorting parameter. Each subsequence of samples yields a statistic that is independent of corresponding statistics yielded by other subsequences of samples. These statistics, each of which corresponds to a subsequence, can then be combined in different ways to derive an estimate of an electrophysiologic response contained in the signal. The presence of non-stationary noise can, to a great extent, be compensated for by appropriately combining the statistics associated with each subsequence.
In one practice of the invention, a plurality of samples of a measured electrophysiologic signal is obtained. The electrophysiologic signal typically includes an electrophysiologic response to a stimulus. The method of the invention seeks to estimate the value of this response.
The method includes defining a plurality of bins, each of which corresponds to a range of values of a sorting parameter associated with each of the samples. Preferably, the range of values for each bin is such that each value of the sorting parameter is associated with at most one bin.
Each sample of the measured signal is then classified into one of the bins on the basis of a value of a sorting parameter associated with that sample. Then, for each bin, a statistic indicative of samples classified into that bin is maintained. On the basis of these bin statistics, the desired electrophysiologic response can then be estimated. In one particular practice of the invention, maintaining the bin statistic includes maintaining a moving average of samples in the bin.
In one practice of the invention, the sorting parameter includes a measure of noise present in the samples. The noise might be electrophysiologic noise, ambient acoustic noise, or any other noise process. The sorting parameter can also be derived from a combination of noise processes.
The estimation of electrophysiologic response can include combining the bin statistics to derive a quantity indicative of the electrophysiologic response. This might include averaging the bin statistics, or evaluating a weighted averaging of the bin statistics, with the weights being manually or automatically selected. In one practice, the weight assigned to a statistic for samples in a particular bin might be indicative of a quality of the samples in the bin. For example, the weight can be inversely proportional to a noise level associated with the particular bin. Alternatively, the weights can be selected to optimize a measure of an extent to which the quantity approximates the electrophysiologic response. The assignment of weights in a weighted average can also include excluding bin statistics associated with particular bins from being considered in evaluating the quantity indicative of the electrophysiologic response.
In another practice of the invention, a sequence of samples is decomposed into a plurality of subsequences, each of which includes samples selected on the basis of a value of a sorting parameter associated with each of the s
Fish & Richardson P.C.
Mallari Patricia C
Massachusetts Eye & Ear Infirmary
Nasser Robert L.
LandOfFree
Measurement of electrophysiologic response does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Measurement of electrophysiologic response, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Measurement of electrophysiologic response will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3269059