A neural complexity measure applied to MEG data in Alzheimer's disease

Objective: A measure of neural complexity (C-N) (Proc. Natl Acad. Sci. USA 91 (1994) 5033) was applied to magnetoencephalography (MEG) data to test the hypothesis that C-N decreases when information processing in the brain is impaired, as is the case in patients with Alzheimer's Disease (AD). Methods: One hundred and fifty-one channel MEGs were recorded in 20 AD patients and 20 healthy age-matched controls in a resting condition with eyes open (EO) and eyes closed (EC). Artifact-free epochs of 117 channels were selected for analysis. C-N and D-2 were computed in different frequency bands, and correlated with the MMSE. Results: The Group X Frequency band interaction was significant for both C-N and D-2. C-N was higher in AD, as compared with controls, in the 2-4 and 4-8 Hz bands, and D-2 was higher in AD patients in the 14-20 and 20-30 Hz bands. The C-N was higher in the EC condition compared to the EO condition, whereas the D-2 was higher in the EO condition. Conclusions: The hypothesis of Tononi et al. (Proc. Natl Acad. Sci. USA 91 (1994) 5033) that the neural complexity decreases in AD patients has to be rejected. However, both neural complexity and the correlation dimension did show differences between controls and AD patients which depended on frequency band. (C) 2003 International Federation of Clinical Neurophysiology. Published by Elsevier Science Ireland Ltd. All rights reserved.