Single unit activity of subthalamic nucleus of patients with Parkinson's disease under local and generalized anaesthesia: Multifactor analysis. (August 2019)

• Record Type: Journal Article
• Authors: Myrov, Vladislav; Sedov, Alexey; Salova, Ekaterina; Tomskiy, Alexey; Belova, Elena
• Is Part Of: Neuroscience research. Volume 145(2019)
• Page Start: 54
• Page End: 61
• Publication Date: 2019-08
• Subjects: Parkinson's disease -- Subthalamic nucleus -- Single unit activity -- Microelectrode recordings -- Propofol -- Data mining; Neurosciences -- Research -- Periodicals; Neurosciences -- Research -- Japan -- Periodicals; Neurology -- Periodicals; Neurosciences -- Periodicals; Neurosciences -- Recherche -- Périodiques; Neurosciences -- Recherche -- Japon -- Périodiques; Neurosciences -- Research; Japan; Periodicals; 612.8
• DOI: 10.1016/j.neures.2018.08.006
• Languages: English
• ISSNs: 0168-0102

Abstract

Highlights: Propofol increased bursting and decreased firing rate in STN single units. Propofol influenced the interplay of parameters of neuronal activity in STN. Random forest algorithm fits better to predict the state of neuronal activity. Abstract: The analysis of neuronal activity in human brain is a complicated task as it meets several limitations, including small sample sizes, dependent variables in the dataset and the short duration of recordings that entangles the analysis procedure. Here, we present the comparative research of neuronal activity in subthalamic nucleus (STN) of 8 Parkinsonian patients undergoing DBS surgery in awake state and under propofol anaesthesia using different statistical approaches. We studied 25 parameters of single unit activity and performed a direct comparison of the parameters between the groups to characterise the changes in STN activity under anaesthesia. We found a significant decrease in firing rate and a prominent increase in bursting of neurons in the anaesthetised state. Also, these data were used to determine the most important parameters for classification. We revealed the differences between parametric and nonparametric approaches regarding the identification of the most important spike train features. The random forest trees algorithm showed a greater accuracy of classification (91.7 ± 1.8%) compared to generalised linear models (82.4 ± 3.8%). The lists of the features important for classification according to F-scores and random forest trees also differed markedly. Our results indicate that feature interactions play a key role in neuronal activity analysis and must be taken into account.