Jack R. Smith

EEG sleep stage scoring by an automatic hybrid system

Abstract The present paper reports the first systematic application of a hybrid sleep EEG scoring system and compares performance of that system with human scoring techniques. The automatic analysis technique can be used on line or 32 times real time, allowing for the scoring of an 8 h sleep EEG record in 15 min. Therefore, the system has obvious potential for reducing the cost and increasing the efficiency of scoring EEG sleep records. Generally, the system has proven to be both accurate and reliable, giving satisfactory levels of agreement upon repeated scoring of the same record or upon scoring different records. The present study indicates further that frequency spectrum analysis is not practical in sleep staging, since the amplitude of delta activity and spindles changes throughout a given night. The incorporation of detectors of phenomena such as K complexes into the sleep-analyzing hybrid computer system will further increase its accuracy and reliability. However, linear analog filters, such as were used for spindle detection, are too sensitive to amplitude variations in the EEG. We have recently developed a non-linear spindle filter, which is not subject to this criticism. This filter will be the subject of a future communication.

Automatic Detection of the K-Complex in Sleep Electroencephalograms

The development and application of an electronic system which is capable of automatically detecting the K-complexes, aperiodic waveforms found in sleep electroencephalograms, is described. The system can be used either on- or off-line.

Ontogeny of delta activity during human sleep.

The study describes the ontogenetic and time-related changes in peak amplitude, frequency, and incidence of delta waves of twenty-five normal subjects between the ages of 3 and 79 years of age. All slow waves greater then 5 muV in magnitude and between 0.5 and 3.0 c/sec were analyzed for the first 6 h of the nights sleep. In this study increasing age was accompanied by decreases in average peak amplitude of delta waves over 5 muV, slowing of delta frequencies, and decreases in incidence of waves greater than 20 muV. Across 2-h epochs of the night there were tendencies for the average peak amplitude of delta waves to decrease, for the average frequency for a given amplitude range to decrease, and for the incidence of low-amplitude delta waves to increase and high-amplitude waves to decrease. The incidence of waves greater than 5 muV remained constant across ages and epochs. Frontal channel delta waves exhibited an age-group difference whereas no such difference was apparent for central channel delta activity, suggesting that the frontal channel may possess advantages over the central because of greater sensitivity to age related changes.

Detection of human sleep EEG waveforms

Abstract A method for detecting waveforms occurring in the human sleep EEG (alpha, beta, delta, sleep spindles, and muscle activity) is presented. Design criteria are provided for the waveform detectors, which can be readily constructed from linear and digital integrated circuits or programmed on a minicomputer.

Automatic Analysis and Detection of EEG Spikes

A method for detecting spikes in the EEG is described. A digital computer is used to perform a detailed analysis of abnormal spikes to obtain quantitative values for spike characteristics and to determine the parameter values to be used in a special-purpose spike detector. This detector is described together with application results.

SAMICOS-A Sleep Analyzing Microcomputer System

A microcomputer-based sleep analyzing system capable of real-time waveform detection and storage is presented. A system design methodology necessary to build the real-time, multiobject, multichannel signal detection scheme is described with the crucial steps in algorithm development and implementation. The system performance is comparable to that of a hybrid, older version of a sleep analyzer which was extensively tested with good results.

Automatic Analysis of Sleep Electroencephalograms by Hybrid Computation

An automated sleep electroencephalogram (EEG) analyzer has been designed and tested in an effort to eliminate tedious and variable human interpretation of experimental EEG data. Data is presented to a hybrid computer from EEG tapes recorded during experimental studies in a human sleep laboratory. Special analog filters are used to identify specific transient waveforms in the EEG. Bandpass filters are used to detect the rhythmical waveforms. The outputs of these filters are then processed by digital logic circuitry, whose algorithms emulate the rules used by human readers quantitating the level of sleep each minute according to the EEG pattem. Preliminary results give 89-percent correlation with a minute-by-minute comparison to the human evaluation of the same test EEG.

A multichannel hybrid system for rapid eye movement detection (REM detection)

Abstract The design, development, and application of a multichannel hybrid system capable of accurate automatic detection of rapid eye movements (REMs) in electrooculograms (EOGs) recorded from sleeping subjects is described. A definition of a rapid eye movement as it appears as a waveform in recorded EOGs is given and used to develop a REM model. This model serves as a basis for the development of the detection system. The interface between the system circuitry and system software is discussed. In addition, comparative results of the application of the system to automated rapid eye movement analysis are presented.

Quantification of abnormal EEG spike characteristics

Abstract A digital computer technique is presented whereby abnormal scalp EEG transients are magnified and processed so that detailed characteristics of their morphology can be precisely quantified. A set of six parameters describing these characteristics is proposed, and data from a quantitative analysis of abnormal spikes chosen from three human epileptics are presented. Finally, the advantages of the presented method are discussed.

Design and Implementation of Linear Phase FIR Filters for Biological Signal Processing

This paper presents a new method for designing digital linear phase, finite impulse response filters with loose frequency response characteristics, but with good time resolution as is required in biological signal conditioning. The design is very simple and has been used with success in the microcomputer implementation of filters for the automated processing of electroencephalographic (EEG) data. Examples and a discussion of possible filter implementations are included.