Objective: To apply pattern recognition algorithms to identify changes in cortical phase amplitude coupling (PAC) responses derived from PD subjects while off and on levodopa/carbidopa drug therapy.

Background: PAC may provide a useful non-invasive electrophysiological surrogate marker of cortical brain activity in PD patients. Eliciting such profiles would have potential in a number of translational applications: a) PD biomarker development; b) PD progression c) Drug and device response profiling; d) Future therapeutic drug/neuromodulation device development to modulate pathological brain activity.

Methods: MEG recordings were acquired from a 306-channel whole head system while an individual with PD was at rest. Recordings and clinical data were acquired from 18 PD subjects both off and on levodopa/carbidopa drug. We used source localization to extract time-varying signals from the following regions of interest (both hemispheres): a) superior frontal gyrus; b) superior frontal sulcus; c) precentral gyrus; d) central sulcus; e) postcentral gyrus. The Direct PAC approach was subsequently used to compute PAC for each region of interest, providing a measure of within-region phase locking of high frequency amplitude (45-280 Hz) with low frequency phase (2-35 Hz). 10 variables (derived from left and right hemispheres) for each of the five regions listed produced a total of 360 individual recordings. Matlab was utilized to create artificial neural network (ANN) models capable of classifying data as either “on” or “off” drug treatment. After randomizing sample order, the samples were placed into training, validation or test sets (30%, 10% and 60%, respectively). Over 150 Boolean classifiers (“Off-drug”/”On-drug”) were developed. Model performance was assessed using the results from the test set only.

Results: One model was capable of accurately assigning 91% of blind test samples (20/22) to their correct class of either being a PAC derived from a patient who was off or on drug treatment.

Conclusions: Preliminary data suggests that ANN may have potential applicability to discern cortical PAC derived from PD patient’s off or on clinical state. Complex pattern recognition may yield useful information regarding dynamic pathogenic neurophysiological mechanisms in PD.

« Back to 2016 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/cortical-phase-amplitude-coupling-pac-identifies-patterns-in-clinical-states-of-parkinsons-disease-pd-using-magnetoencephalography-meg/