EEG source extraction by autoregressive source separation reveals abnormal synchronization in Parkinson's disease

• Published in: Conference proceedings (IEEE Engineering in Medicine and Biology Society. Conf.) Vol. 2009; pp. 1868 - 1872
• Main Authors: Chiang, J., Wang, Z.J., McKeown, M.J.
• Format: Conference Proceeding Journal Article
• Language: English
• Published: United States IEEE 01.01.2009
• Subjects: Algorithms; Brain - physiology; Brain - physiopathology; Brain Mapping - methods; Brain modeling; Clustering algorithms; Electroencephalography; Electroencephalography - methods; Gaussian distribution; Humans; Independent component analysis; Likelihood Functions; Models, Neurological; Parkinson Disease - physiopathology; Parkinson's disease; Reference Values; Regression Analysis; Scalp; Signal Transduction; Software; Source separation; USA Councils; Volume measurement
• ISSN: 1094-687X; 1557-170X
• DOI: 10.1109/IEMBS.2009.5332613

Recent research efforts in studying brain connectivity has provided new perspectives to understanding of neurophysiology of brain function. Connectivity measures are typically computed from electroencephalogram (EEG) signals, yet the presence of volume conduction makes interpretation of results difficult. One possible alternative is to model the connectivity in the source space. In this study, we proposed a novel source separation technique in which EEG signals are represented as a state-space framework. The framework jointly models the underlying brain sources and the connectivity between them in the form of a generalized autoregressive (AR) process. The proposed technique was applied to real EEG data collected from normal and Parkinson's patients during a motor task. The extracted sources revealed the abnormal beta activity in Parkinson's subjects and showed similar biological networks as previous studies.