Introduction

This book describes tools for the quantitative analysis of the EEG. Here the term EEG includes EEG, EMG, and EOG channels. Although polysomnography also includes much lower frequency channels such as respiration and oxygen saturation, techniques for their quantitative analysis are not discussed here. My goal is to introduce various techniques and to compare the results obtained from different techniques. I developed, applied, and published these tools during, what turns out to be, almost forty years of research. Many of the described tools were once difficult to apply, but this is no longer the case as giant steps in computer hardware and software make it possible to provide easily applicable quantification tools. The most frequently used techniques are based on a data reduction of frequency data obtained with the Discrete Fourier Transform (DFT). The popularity of quantitative techniques based on the DFT is, in my opinion, in large part due to its widespread availability rather than any significant findings. Like most tools, DFT analysis can be improperly applied. This book first attempts to explain quantitative techniques based on the DFT and provide guidelines for improving the accuracy of the results. Then quantitative techniques in time domain are talked about. After that data obtained by the two techniques are compared side by side in order to illustrate the differences in the methodologies.

A research tool, in addition to providing objective information, needs to be widely disseminated. Research data is not objective until the methodology used to obtain the data is clearly described and the research results replicated. The quantitative techniques described here are all contained in the polysomnography software package Polysmith. The methodology is currently proprietary, although the basic methodologies and criteria are described in the published literature. In an effort to expand sleep research, Neurotronics is willing to make the techniques described here available to investigators in addition to those already using the Polysmith software, either by providing a software module which can be incorporated into other commercial software or by providing detailed waveform detection criteria. The graphing and database software is also proprietary, but that software is not necessary for the use of the quantitative tools.

There are many publications that have used the techniques described here, but the published literature is not reviewed in this document. A list of my publications relative to the current discussion is included as an appendix. Another appendix describes the process of acquiring a channel of EEG data. Specifications and evaluation procedures are included for anyone who might be interested.

As for my qualifications to make some of the following comments, I have the advantage of working in the field for forty years; offsetting this advantage is that I have forgotten too much. I have founded two companies which produced products for the PSG market and I designed both hardware and software for this market. I also spent thirty years as an electrical engineering professor, where my teaching and research included electronics, filter design, and both analog and digital signal processing. During that time I had the great fortune and pleasure to spend sabbatical leaves working on quantitative EEG analysis at the Centre Nationale Researche Scientific in Marseilles, France; The Medical school of Geneva, Geneva, Switzerland; and the Tokyo Metropolitan Research Institute. Investigators from all of these institutions contributed to my research results as did many graduate students working in my laboratory and several colleagues at Neurotronics, especially Mr. Hemanth Pedddaneni, who has made many technical contributions and has managed the editing of this book. My work on quantitative EEG analysis was greatly enhanced by a long-term collaboration with Dr. Ismet Karacan, who welcomed this engineer into his research group forty years ago, pointed me in the right direction, and raised the bar. It is ironic that the quantitative tools I developed are not well known, but our Polysmith software is a huge success. Perhaps this booklet will encourage research investigators to use the many research tools described here.

Jack R. Smith

Neurotronics Incorporated

October 14, 2006