Aiding the detection of Alzheimer's disease in clinical electroencephalogram recording by selective de-noising of ocular artifacts

Abstract

Alzheimer's disease is a progressive, irreversible, neurodegenerative disease with no known cause or cure. Clinical recording of the electroencephalogram (EEG) Signal for patients with Alzheimer's disease has shown a characteristic increase in slow quantitative EEG (qEEG) frequencies and a smaller decrease in the fast activities. A commonly encountered problem in clinical practice during EEG recording is the blanking of the EEG signal due to blinking or movements of the user's eyes. Recent research on the effectiveness of the various techniques for filtering these ocular artifacts has shown that, while a significant portion of the EEG data is lost, there is also some remnant artifact subsequent to the de-noising process. Considering these aberrations, along with the fact that most of these methods require continuous monitoring of the electrooculargram (EOG) signal as well, prompted us to use Haar wavelets to accurately detect the presence of ocular artifacts and de-noise them. The remarkable performance of this technique over conventional methods guided us to extend it to clinically recorded Alzheimer's EEG signals as well. This paper describes the use of Haar wavelets for selective detection and de-noising of the ocular artifacts in clinically recorded Alzheimer's EEG signals and discusses both the errors involved as well as the drastic improvements in efficiency over conventional techniques.