Advanced time-series analysis of MEG data as a method to explore olfactory function in healthy controls and Parkinson's disease patients

Abstract

Objectives: To determine whether time-series analysis of magnetoencephalography (MEG) data is a suitable method to study brain activity related to olfactory information processing, and to detect differences in odor-induced brain activity between patients with Parkinson's disease (PD) and controls.

Methods: Whole head 151-channel MEG recordings were obtained in 21 controls and 20 patients with PD during a 10-min olfactory stimulus paradigm, consisting of 10 alternating rest-stimulus cycles (30 s each), using phenylethyl alcohol administered by means of a Burghart olfactometer. Relative spectral power and synchronization likelihood (SL; an unbiased measure of functional connectivity) were calculated for delta, theta, alpha1, alpha2, beta, and gamma frequency bands.

Results: In controls, olfactory stimulation produced an increase in theta power and a decrease in beta power. In patients with PD, there was a decrease in alpha1 power. No significant interaction between group and condition was found for spectral power. SL analysis revealed a significantly different response to olfactory stimulation in patients with PD compared to controls. In controls, the odor stimulus induced a decrease in local beta band SL. The response in patients with PD involved a decrease in intrahemispheric alpha2 band SL.

Conclusion: This is the first study to show that time-series analysis of MEG data, including spectral power and SL, can be used to detect odor-induced changes in brain activity. In addition, differences in odor-induced brain activity were found between patients with PD and controls using analysis of SL, but not of spectral power.

Figure 1

A schematic representation of the olfactory stimulus protocol. Phenylethyl alcohol (PEA, 40% v/v) was delivered for 1 s every 4s during a 30 s period in the “stimulus” condition. During the 30 s “rest” period, subjects received odorless air. A total of 10 alternating rest‐stimulus cycles were presented. Four seconds of odorless air preceded the first stimulus condition; MEG recordings acquired during these four seconds were not used in the analyses.

Figure 2

Sensor clustering and selection of relative spectral power and synchronization likelihood (SL) measures. (A) Clustering of MEG sensors above major cortical areas; midline sensors were excluded from spectral power and (SL) analysis. (B) Schematic representation of regions of interest (ROIs) used to calculate spectral power and short‐distance local SL. (C) Long‐distance interhemispheric connections used to calculate SL. (D) Long‐distance intrahemispheric connections used to calculate SL. Arrows indicate SL connections used. L, left; R, right; F, frontal; C, central; P, parietal; O, occipital; T, temporal.

Figure 3

Percentage of change and standard errors of the mean in relative spectral power (stimulus compared to rest condition), for each frequency band. Gray bars represent control subjects, white bars represent patients with PD. * indicates P‐value <0.05, when comparing all rest and stimulus epochs in a multilevel statistical model.

Figure 4

Percentage of change and standard errors of the mean in synchronization likelihood (stimulus compared to rest condition), for each frequency band. (A) Local synchronization likelihood. (B) Interhemispheric synchronization likelihood. (C) Intrahemispheric synchronization likelihood. Gray bars represent control subjects, white bars represent patients with PD. * indicates P‐value <0.05, when comparing all rest and stimulus epochs in a multilevel statistical model. # indicates P‐value <0.05 for the group × condition interaction in a multilevel statistical model.