James W. Ward

Autoregressive Estimation of Short Segment Spectra for Computerized EEG Analysis

The hypothesis that an electroencephalogram (EEG) can be analyzed by computer using a series of basic descriptive elements of short duration (1-5 s) has prompted the development of methods to extract the best possible features from very short (1 s) time intervals. In this paper, the merits of three alternative methods for estimating spectral features are compared to the fast Fourier transform (FFT). These procedures, based on autoregressive (AR) modeling are: 1) Kalman filtering, 2) the Burg algorithm, and 3) the Yule-Walker (YW) approach. The methods are reportedly able to provide high resolution spectal estimates from short EEG intervals, even in cases where intervals contain less than a ful period of a cyclic waveform. The first method is adaptive, the other two are not. Using Akaikes final prediction error (FPE) criterion, it was demonstrated that a fifth-order filter is sufficient to estimate EEG characteristics in 90 percent of the cases. However, visual inspection of the resulting spectra revealed that the order indicated by the FPE criterion is generally too low and better spectra can be obtained using a tenth-order AR model. The Yule-Walker method resulted in many unstable models and should not be used. Of two remaining methods, i.e., Burg and Kalman, the first provides spectra with peaks having a smaller bandwidth than the Kalman-flter method. Additional experiments with the Burg method revealed that, on the average, the same results were obtained using the FFT.

Sleep patterns in a patient with a brain stem infarction involving the raphe nucleus

Abstract A patient with an infarction in the basis pontis had a decreased sleep time but a normal percentage of REM sleep. Autopsy revealed a well defined infarction destroying the pontine and midbrain portions of the raphe nucleus. This patients polygraphically recorded sleep patterns are similar to the sleep patterns of experimental animals with raphe lesions.

Quantitative assessment of the electroencephalogram in renal disease

EEGs wre recorded from renal patients to determine if there are quantifiable characteristic changes in the EEG was quantified by calculating the percentage of spectral power in the bandwidth 3-7 c/sec referrred to a frequency range of 3-13 c/sec and by computing the mean frequency of the dominant rhythm in the EEG. Blood urea nitrogen and creatinine concentrations, as well as a self-assessment of the patients clinical condition, were recorded. The general finding of this research is that EEG slowing, as evaluated by power spectral techniques, is correlated with uremia-associated variables. 1. In a non-dialyzed patient population with renal failure, slowing in the EEG was found to be directly corelated with increased creatinine concentrations. 2. Quantitative measures of slow wave activity computed using power spectral techniques were found to be highly corelated with an estimate of slowing made by an electroencephalographer. 3. Compared with undialyzed azotemic patients, malignant hypertensive patients with comparable serum creatinine concentrations typically displayed increased slow wave activity, while slowing was generally reduced in the dialyzed patient population. 4. A series of EEGs recorded from one patient during the first three dialyses of her life revealed that slow wave activity decreased during each successive dialysis. In another patient, all quantified EEG values recorded prior to renal transplantation significantly improved after transplantation...

Visually evoked cortical potentials in renal failure: transient potentials.

Transient visually evoked cortical potentials (VECPs) were recorded from patients with renal disease. Changes in VECP latencies are described for undialyzed patients, patients receiving dialysis therapy, and patients who received kidney transplants. Characteristics of VECP latencies in these patient groups as well as examples of changes in latency values with respect to time for two individual patients are examined. The basic overall finding is that the VECP latencies increase as a patients clinical condition deteriorates and normalize as the condition improves.

Artificial intelligence methods in quantitative electroencephalogram analysis

Abstract A set of programs designed to implement artificial-intelligence-related concepts in quantitative electroencephalogram analysis are described. The programs use rule-based logic with top down parsing (backward chaining) to evaluate EEG data. A simple implementation of fuzzy logic in premise clauses of ‘IF-THEN’ rules is included.

The recording of the fetal electrocardiogram

Abstract An improved method for recording the fetal electrocardiogram is described; it uses the 3-channel electroencephalograph and crystograph ink writers. Fetal waves were recorded successfully from the seventeenth week of pregnancy onward, and 82 per cent of the tracings taken during this period were positive. Certain advantages and uses of the method are indicated.

The EEG Analysis System of the National Cooperative Dialysis Study

As a renal patients clinical condition deteriorates, slow-wave-related activity typically appears in the electroencephalogram (EEG) and disappears or is reduced as the clinical situation ameliorates. These changes can be quantitatively monitored by computer calculation of spectral estimates of the amount of slow-wave-related power in the EEG. The computer quantified EEG was used as one end point measure in the study of patients who participated in the National Cooperative Dialysis Study (NCDS). This study was designed to evaluate quantitative guidelines for administering dialysis therapy on a individualized basis.

Quantitative assessment of photic driving in renal failure

Photically driven EEGs were recorded from patients with renal disease using photic stimulation at integer rates between 3 and 12 flashes/sec. Changes in the structure of the power spectrum of the potentials produced by this stimulus paradigm are described as a function of a patients clinical state. The basic overall findings are that (1) harmonic activity is attenuated while activity below the fundamental driving frequency is increased as a patients clinical condition deteriorates, and (2) that these effects are substantially reversed and controlled by means of dialysis and renal transplantation.

Computer Quantification of Electroencephalographic Data Recorded from Renal Patients

Abstract Computer quantification of EEG records recorded from patients with renal disease reveals a slowing of the EEG correlated with deterioration of the patients clinical condition. Electroencephalographic records were analyzed (1) by time averaged autocorrelation and subsequent Fourier transformation and (2) by generation of compressed spectra arrays (CSA). Quantification of the CSA provided estimates of the mean and variance of the percent power in selected frequency bandwidths in the EEG. This quantification has proved quite useful in the detection of slowing in EEG records.

Identification and labeling of EEG graphic elements using autoregressive spectral estimates

Syntactic EEG analysis requires descriptive labeling of short (1 s) epochs in an EEG. While discriminant analysis methods are useful for this purpose, significant improvement in label correctness can be achieved using the heuristic method described in this paper. The method is based on the estimation of frequency spectra by autoregressive (AR) modeling. The location of the peak frequencies and the power of these peaks are used to assign labels to 1 s epochs. Artefacts and epochs with exceptionally high or low amplitude and/or frequency values are identified as well. The assignment of labels is accomplished by comparing parameters, extracted from the power spectra estimated for 1 s epochs, with thresholds. These thresholds are automatically adapted to each individual EEG lead. In this paper, the method is outlined and its performance is compared with a discriminant analysis approach and visual labeling.