SPARSE BUMP MODELING OF MILDAD PATIENTS: Modeling Transient Oscillations in the EEG of Patients with Mild Alzheimer's Disease

摘要

We explore the potential of bump modeling to extract transient local synchrony in EEG, as a marker for mildAD (mild Alzheimer's disease). EEG signals of patients with mildAD are transformed to a wavelet time-frequency representation, and afterwards a sparsification process (bump modeling) extracts time-frequency oscillatory bursts. We observed that organized oscillatory events contain stronger discriminative signatures than averaged spectral EEG statistics for patients in a probable early stage of Alzheimer's disease. Specifically, bump modeling enhanced the difference between mildAD patients and age-matched control subjects in the θ and β frequency ranges. This effect is consistent with previous results obtained on other databases.