Richard H. Jones’s research while affiliated with University of Denver and other places

Response criteria for a significant averaged evoked heart rate response to 63 and 80 dB SPL tones at 500, 1000, 2000, 4000, 6000, and 8000 Hz were objectively defined by means of autoregressive analysis in 25 newborns (Group I). This statistical procedure for detecting significant heart rate (HR) change was based on predictions derived from heart rate activity occurring prior to onset of a stimulus. The Group I criteria were applied to HR responses of 65 newborns (Group II) averaged over 2 to 24 trials. Averaged HR changes were low for 2 trials but increasingly higher when based on the average over all 63dB SPL tones, all 80dB SPL tones, and finally, all tones; 43, 65, and 77 percent of the infants responded, respectively. For the last three conditions, the averages were based on 12, 12, and 24 trials, respectively. Results suggest that in human newborns the averaged HR response may (a) function as a reliable auditory response measure, (b) supplement other measures of neonatal responsivity to suprathreshold tones, (c) provide information on early auditory abilities and the developmental course of auditory functioning, and (d) serve as a technique for investigating cardiac orienting and the effects of behavioral teratogens on auditory functioning.

This report presents an autoregressive technique for detecting statistically significant changes in brain activity to tones. The change detection model is applied to stationary time series electroencephalogram samples from sleeping newborn infants. The electroencephalic responses of neonates to tones are quantified and analyzed in terms of t-statistics. Confidence limits applied to averaged t-statistics objectively and reliably defined statistically significant late components in newborns.

In order to examine the reliability and validity of classification of EEG maturational levels, brain activity before and after peripheral sensory stimulation was measured in two groups of full-term newborn infants (N=50 and 48). Autoregressive coefficients from two derivations (C-A, O-A) were classified by means of logistic discrimination relative to three age groups: a low birthweight group of less than 35 weeks gestational age, a comparable two-day-old group, and a group 6 weeks to 3 months of age. Results demonstrated that EEG maturational level of two-day-old full-term neonates can be reliably and validly classified. Autoregressive spectral analysis may function as a major technique for the hypothesis testing and classification of infant EEG.

Spectral analyses show unilateral photic driving in newborn human infants to bilateral repetitive visual stimulation. Results are interpreted as evidence of dominance in the right hemisphere for rhythmic visual stimuli and lack of interhemispheric integration.

Without Abstract

An adaptive method for detecting change in digitized cardiotachometer recordings has been developed which takes into account the nonstationary statistical structure of the data. The digitized data are smoothed to reduce the variance at high frequencies caused by discontinuities inherent in cardiotachometer output. Based on a first-order autoregression, which has been shown to be appropriate for heart rate data, the adaptive procedure uses estimates of the parameters which are most influenced by the recent observations. Decreasing weight is given to past prestimulus data, and the estimates are updated with each stimulus. A test for change is then applied to the poststimulus regions at each time point, yielding a t statistic. The t 2 's can then be averaged to give a test for change over an interval.

A method is presented for detecting change in biological multivariate stationary processes of a single subject after stimulation. A finite multivariate autoregression is fitted to pre-stimulus data using a step-wise procedure with tests of significance. The fit of the model is also checked by comparing the estimated spectra, phase, and coherence with fitted curves. The statistic which tests for change at a given time point is a quadratic form involving the one-step prediction error vector and the inverse of the one-step prediction error covariance matrix. Under the hypothesis of no change, and for a Gaussian process, these statistics have independent chi-square distributions. The technique has been applied to the detection of change in the brain waves of two human newborn infants following stimulation.

An auto-regression model using a step-wise procedure for estimating the parameters has been developed to test for change in the post-stimulus onset EEG of a single subject. A stationary process is assumed for the pre-stimulus data and if there is no significant change, for the post-stimulus region. Change can be detected at any point in the time series giving a measure of response latency and duration.This analysis was performed utilizing the responses of a human newborn to sensory stimulation to demonstrate the applicability of the model for detecting changes in digitized EEG data and relating these to underlying neurophysiological activity.RésuméUn modèle d'auto-régression utilisant un déroulement par étapes pour l'estimation des paramètres a été mis au point afin de tester tout changement dans les parties de l'EEG immédiatement consécutives à un stimulus sur un sujet isolé. Les données pré-stimulus sont considérées comme un processus stationnaire et, s'il n'y a pas de modification significative, il en est de même de la période post-stimulus. Une modification peut être décelée à tout point de la série temporelle qui donne une mesure de latence et de durée de la réponse.Cette analyse a été réalisée en utilisant les réponses d'un nouveau-né humain à la stimulation sensorielle pour démontrer l'applicabilité du modèle à la détection de changements dans les données EEG quantifiées et à leur mise en relation avec l'activité neurophysiologique sousjacente.