Surgery – Diagnostic testing – Ear or testing by auditory stimulus
Reexamination Certificate
1999-04-26
2001-03-13
O'Connor, Cary (Department: 3736)
Surgery
Diagnostic testing
Ear or testing by auditory stimulus
C600S544000
Reexamination Certificate
active
06200273
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to the field of assessing hearing capacity in humans. More particularly, the invention is a method for quantifying the probability that an auditory brainstem response (ABR) is present in an electrophysiologic recording from a human infant.
2. Background of the Invention
The ABR is a waveform of fluctuating electrical potential over time, which may occur in response to a brief, transient acoustic stimulus such as a click. The ABR originates in the neurons of the auditory nerve and its higher connections in the brain stem. When recorded from electrodes on the scalp or neck, it is less than one microvolt in size, and is obscured by much larger ongoing random potentials that arise elsewhere in the brain and the musculature of the head and neck. Computer summation or averaging of the responses to several thousand stimuli presented at rates typically in the range of 20-50 per second is required to enhance the ABR “signal” relative to the background electrical “noise”, and to render it visually detectable in the summed or averaged response.
The presence or absence of an ABR for a specific type and intensity of stimulus can be used as a proxy for overt behavioral response, indicating whether or not the stimulus was audible. This is the basis for an electrophysiologic hearing screening test, of particular value in subjects such as infants who are unable to give reliable, behavioral responses to sound.
In the newborn population, it is widely acknowledged that it is important to detect and manage hearing loss as early as possible, and preferably in the first six months, to facilitate development of speech, language and cognitive skills. In 1993, the National Institute on Deafness and Other Communication Disorders sponsored a Consensus Conference on Early Identification of Hearing Impairment in Infants and Young Children. That conference recommended screening for identification of hearing impairment in the newborn period for all infants regardless of the presence or absence of risk factors for hearing loss, that is, universal infant hearing screening. These recommendations were endorsed and reiterated soon after by the American Academy of Audiology and the Joint Committee on Infant Hearing, 1984. Many states have recently implemented, or are in the process of implementing, such screening programs. This widespread endorsement of mass hearing screening of neonates and infants has created a challenge for scientists and clinicians to have fast and accurate tools ready for evaluating potential hearing loss in infants.
ABR testing is well established as a core part of most screening protocols. The clinical utility of ABR-based hearing screening tests depends critically on the accuracy of the ABR detection decisions. Such decisions are intrinsically prone to error, because they involve the detection of a signal in random noise that may obscure a genuine signal or masquerade as a signal when none is present. False-positive ABR detection leads to a false-negative screening test: the hearing-impaired child passes the screening test and receives no intervention. Other manifestations of disorder may be ignored, given that the test was passed, so the screening does active harm. False-negative ABR detection tests cause false-positive screening tests; this precipitates needless follow-up diagnostic assessment costs, as well as indirect costs of mislabeling a normal child.
A distinction must be drawn between detection tests that are empirical and those that are analytic. Empirical tests are based upon experimental studies of the distributions of a given test statistic when response is thought to be present or absent. Usually the determination of response presence or absence is based on expert subjective assessment of the average records obtained in a set of subjects. There are two major difficulties with this approach. First, the expert judgments may be wrong, which clearly confounds the assessment of the accuracy of the test statistic. Second, there is no proof that the results observed in one set of subjects will necessarily apply to a different set of subjects or to a situation in which any feature of the data recording or analysis is changed. This is a failure of generalizability of the empirical validation process.
Analytic methods, in contrast, do not appeal to experimental validation datasets. They are based upon known properties of known statistical distributions relating to the chosen test statistic. Thus, rather than relying on empirical experimental data, analytic methods capitalize upon the vast body of statistical distribution theory and statistical tables of distributions. It is necessary to show that real data satisfy certain assumptions that are required for certain distributions to pertain, but these assumptions may be weak, easily satisfied, and easily proven to hold. Such methods are both highly quantitative, yielding known and specifiable rates of decision error, and are also highly generalizible across datasets and measurement conditions.
A crucial characteristic of a good statistical response detection test is that it has the highest possible statistical power. Power is the probability that the test will correctly detect a response that is genuinely present. Less than optimal test power is very disadvantageous in practical terms. A loss of power translates directly to longer test time than necessary to reach the statistical criterion for response detection. This is a major practical disadvantage because some babies yield satisfactory measurement conditions for only brief periods of time, they may be untestable due to test inefficiency. Also testable babies will take longer to test than necessary which increases costs and decreases throughput. This factor will be especially crucial in light of the implementation of universal newborn hearing evaluation protocols now mandated in many states. Third, the use of a test that is less powerful than necessary will result in larger rates of detection decision error than would be possible with a more powerful test.
Prior Art Detection Systems
Current approaches to automated detection of ABRs include techniques that evaluate the time-domain waveform and those that assess spectral characteristics (frequency domain). Automated detection of neonatal click-evoked ABR to low-level stimuli for mass screenings have primarily involved analysis in the time domain, although one known system includes both time and frequency domain analysis. At present, four systems have been used or sold as “automated infant ABR screening” devices. By that we refer to those devices in which decisions regarding ABR presence or absence or test “Pass” or “Fail” (sometimes called “Refer”) is made by the system itself (not by the examiner) based on some predetermined criteria that are discussed below.
The general approach of the detection algorithm employed by the most commonly used system for automated ABR detection in infants appears to be as follows: A set of sample points are weighted according to their relative magnitude in the standard infant ABR waveform. It is not clear how the position or number of the data points are selected. The polarities of the amplitude of each point in a standard or template are compared with those observed at the corresponding latency in each sweep during averaging. Each time a sweep is sampled, the correspondence of polarity between the data and the template at each of the selected time points yields a count of +1. After every 500 sweeps, the template points are shifted in increments of 0.25 ms over a 3 ms range to locate the position of maximum polarity correspondence. Presumably, this is done using an accumulated average of some kind, but this is not clear. Each sample in each sweep constitutes a trial and running counts of the numbers of polarity matches and trials are accumulated. Because the probability of a polarity match for each point is 0.5 if the response is absent, a quantitative hypothesis test can be constructed based on a binomial
Hyde Martyn
Sininger Yvonne S.
Blakely , Sokoloff, Taylor & Zafman LLP
House Ear Institute
Marmor II Charles
O'Connor Cary
LandOfFree
Power-optimized cumulative, sequential statistical method... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Power-optimized cumulative, sequential statistical method..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Power-optimized cumulative, sequential statistical method... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2450848