A method for early prediction of parkinson's disease freezing gait based on electroencephalogram high-order phase characteristics

Abstract

The application discloses a Parkinson's disease freezing gait early prediction method based on electroencephalogram high-order phase characteristics. The method comprises the following steps: acquiring scalp electroencephalogram signals of multiple target brain regions of a user and performing pretreatment; applying a surface Laplace space filtering algorithm to the pretreated signals to suppress volume conduction effect; extracting multi-dimensional neuroelectrophysiological characteristics from the filtered signals in real time, wherein the characteristics at least include phase locking values of target brain region networks in frequency bands and phase-amplitude coupling strength indexes in target brain regions; inputting the multi-dimensional characteristics into a pre-trained machine learning prediction model to output a probability quantitative value of physical freezing gait of the user in future seconds. By introducing surface Laplace filtering and high-order cross-frequency coupling characteristics, the application effectively improves signal space resolution and specificity of prediction characteristics, realizes low-delay and high-accuracy early prediction, and can be used for wearable electroencephalogram devices.

Images