Automated Quantification of Human Alpha Rhythm

Abstract

This thesis seeks to quantify human alpha rhythm in order to obtain better measures to test theoretical models of neurodynamics, with potential clinical applications for the method of identification. An automated algorithm is developed in Chapter 2 to facilitate collection of objective data from an expanded alpha band (4–14 Hz) in a large number of subjects. This method avoids subjective bias inherent to traditional visual identification of the alpha activity, produced multiple peak information (if multiple peaks exist) that is absent in qEEG measures, and uses information from multiple electrode sites to eliminate spurious peaks. This method is tested and validated on 100 subjects. In addition to traditional measures of alpha activities such as the frequency and amplitude, further measures were devised to better quantify the alpha rhythm and its spatial characteristics. Background spectra in the alpha range are also characterized. In Chapter 3 the algorithm is applied to a large (1498 subjects) database of normal healthy subjects of approximately equal number in each sex, as well as a large span in age (6–86 years), in order to establish typical parameter ranges. Analysis is done comparing the age and the topological trends that are known variants in the alpha rhythm. Investigations are also performed to test for potential sex differences and/or lateralities. Key results are that double alpha peaks are resolved in a large proportion of the subjects ( 50%), while single alpha peak cases are likely to be double-peak cases in which the two peaks are not resolved. Age trends in measures of alpha activity show increase of alpha frequency from childhood to adolescence, a plateau in adulthood, and a slight decline in old age; a decrease in alpha amplitude in old age is also observed. These findings are consistent with previous literature and provide better statistics. Topological distribution of the alpha activity on the head is also explored, with little lateral asymmetry observed. No statistically significant differences between the sexes are found. The C++ code that was developed and utilized in this thesis is included in Appendix B. It is provided on disk and is available online. A study carried out on the same group of subjects to determine age-related trends of EEG parameters produced by model fitting is included in Appendixes C, to provide a comparison between the methods and highlights corresponding results.