Stanford Research Communication Program
  Home   Researchers Professionals  About
 
Archive by Major Area

Engineering
Humanities

Social Science

Natural Science

Archive by Year

Fall 1999 - Spring 2000

Fall 2000 - Summer 2001

Fall 2001 - Spring 2002

Fall 2002 - Summer 2003


 

 

 


Reading Brainwaves with Less Pain and Hassle

Tim Uy
Applied Physics
Stanford University
November 2002


Neurologists and experimental psychologists study electrical activity within the brain by recording voltage fluctuations at the scalp. This is called electroencephalography or EEG. In conventional or "wet" EEG, scalp abrasion and use of electrolytic paste are required to insure a good electrical connection between sensor and skin. With repeated application, abrasion can become painful to the patient. This limits the number of sessions a patient can undergo. Recently, several groups have produced "dry" sensors which do not require abrasion or conductive paste. These dry sensors also reduce setup time from about 30 minutes to less than 30 seconds. I am designing methods to verify that dry EEG sensors perform just as well as wet EEG sensors. My methods can be divided into two categories: single-sensor studies and multi-sensor studies.

New types of EEG sensors are usually characterized using single-sensor studies. In one such study, noise from the sensor can be recorded and summarized by computing its spectrum. The spectrum is a measure of how a signal's power is distributed by frequency. The spectrum of a beam of light, for instance, can be seen by passing it through a prism. In EEG, certain frequencies are more important than others. The computed spectrum allows us to make sure that noise at those frequencies does not overwhelm the true EEG signal. Another oft-implemented study is a simultaneous recording of brain waves from both wet and dry sensors in close proximity. Because the recording location is not exactly the same, rather than a mathematical comparison, the analysis typically ends with a visual comparison between the two resulting waves.

While these conventional methods provide some measure of quality, a definitive claim on the feasibility of dry sensors requires more quantitative tests. A significant portion of my research involves modifying conventional single-sensor studies to be more quantitative and creating new single-sensor studies as necessary. For example, rather than simply presenting the noise spectrum for the dry sensor, I overlay it with the spectrum of a wet sensor recorded using the same equipment and methodology. This allows a numerical comparison between the sensors. In a typical recording, noise can come not just from the sensor, but from the cables, the connection to the recording equipment, and the environment. Since noise sources external to the sensor can never be completely removed, the recording conditions should be matched as closely as possible for dry and wet.

For the simultaneous recordings, quantitative comparisons are not justified because the brain waves at the two different locations will be slightly different. To address this issue, I make simultaneous recordings from two wet sensors placed at the two locations. I take the difference of these two recordings and compute the spectrum of the resulting wave. I repeat the process, replacing one of the wet sensors with a dry sensor. By comparing these two spectrums, I can show quantitatively if there are significant differences between wet versus wet, and dry versus wet.

In addition to these improvements to the single-sensor studies, I have found that it is necessary to include multi-sensor studies in order to compare the performance of wet and dry EEG in a realistic environment. These multi-sensor studies compare results from EEG experiments using wet sensors and the same experiment using dry sensors. I am working on developing studies based on EEG experiments with well accepted results. Consistency in the results will show that dry EEG is an acceptable substitution for wet EEG. The claim is often made by developers of dry sensors that single-sensor studies are sufficient. However, this assumes all sensors will behave in the same manner. In single-sensor studies of dry sensors from one particular group, I discovered that signal quality varied quite widely from sensor to sensor. Thus, it is necessary to qualify individual dry sensors using the single-sensor studies. This qualification is then checked using multi-sensor studies.

Accuracy of the EEG is paramount in the medical diagnosis of epilepsy and in experimental psychology,. My single-sensor and multi-sensor studies will show whether the current generation of dry sensors can replace wet sensors. These methods can also be used as standardized tests for any researchers developing new EEG sensors. Someday dry EEG will be widely used. When this happens, it will reduce the cost of EEG by virtually eliminating setup and cleanup time. More patients can come in more frequently which will greatly increase the volume of data and hopefully our understanding of the human brain.