Prediction of mild parkinsonism revealed by neural oscillatory changes and machine learning

Machine Learning Identifies a Rat Model of Parkinson's Disease via Sleep-Wake Electroencephalogram

Alpha-synuclein induced degeneration of the midbrain substantia nigra pars compact (SNc) dopaminergic neurons causes Parkinson's disease (PD). Rodent studies demonstrate that nigrostriatal dopamine stimulates pallidal neurons which, via the topographical pallidocortical pathway, regulate cortical activity and functions. We hypothesize that nigrostriatal dopamine acting at the basal ganglia regulates cortical activity in sleep and wake state, and its depletion systemically alters electroencephalogram (EEG) across frequencies during sleep-wake state. Compared to control rats, 6-hydroxydopamine induced selective SNc lesions increased overall EEG power (positive synchronization) across 0.5-60 Hz during wake, NREM (non-rapid eye movement) sleep, and REM sleep. Application of machine learning (ML) to seven EEG features computed at a single or combined spectral bands during sleep-wake differentiated SNc lesions from controls at high accuracy. ML algorithms construct a model based on empirical data to make predictions on subsequent data. The accuracy of the predictive results indicate that nigrostriatal dopamine depletion increases global EEG spectral synchronization in wake, NREM sleep, and REM sleep. The EEG changes can be exploited by ML to identify SNc lesions at a high accuracy.

Parkinsonian daytime sleep-wake classification using deep brain stimulation lead recordings

Excessive daytime sleepiness is a recognized non-motor symptom that adversely impacts the quality of life of people with Parkinson's disease (PD), yet effective treatment options remain limited. Deep brain stimulation (DBS) of the subthalamic nucleus (STN) is an effective treatment for PD motor signs. Reliable daytime sleep-wake classification using local field potentials (LFPs) recorded from DBS leads implanted in STN can inform the development of closed-loop DBS approaches for prompt detection and disruption of sleep-related neural oscillations. We performed STN DBS lead recordings in three nonhuman primates rendered parkinsonian by administrating neurotoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP). Reference sleep-wake states were determined on a second-by-second basis by video monitoring of eyes (eyes-open, wake and eyes-closed, sleep). The spectral power in delta (1-4 Hz), theta (4-8 Hz), low-beta (8-20 Hz), high-beta (20-35 Hz), gamma (35-90 Hz), and high-frequency (200-400 Hz) bands were extracted from each wake and sleep epochs for training (70% data) and testing (30% data) a support vector machines classifier for each subject independently. The spectral features yielded reasonable daytime sleep-wake classification (sensitivity: 90.68 ± 1.28; specificity: 88.16 ± 1.08; accuracy: 89.42 ± 0.68; positive predictive value; 88.70 ± 0.89, n = 3). Our findings support the plausibility of monitoring daytime sleep-wake states using DBS lead recordings. These results could have future clinical implications in informing the development of closed-loop DBS approaches for automatic detection and disruption of sleep-related neural oscillations in people with PD to promote wakefulness.

Neuronal oscillations predict deep brain stimulation outcome in Parkinson's disease

BACKGROUND

Neuronal oscillations are linked to symptoms of Parkinson's disease. This relation can be exploited for optimizing deep brain stimulation (DBS), e.g. by informing a device or human about the optimal location, time and intensity of stimulation. Whether oscillations predict individual DBS outcome is not clear so far.

OBJECTIVE

To predict motor symptom improvement from subthalamic power and subthalamo-cortical coherence.

METHODS

We applied machine learning techniques to simultaneously recorded magnetoencephalography and local field potential data from 36 patients with Parkinson's disease. Gradient-boosted tree learning was applied in combination with feature importance analysis to generate and understand out-of-sample predictions.

RESULTS

A few features sufficed for making accurate predictions. A model operating on five coherence features, for example, achieved correlations of r > 0.8 between actual and predicted outcomes. Coherence comprised more information in less features than subthalamic power, although in general their information content was comparable. Both signals predicted akinesia/rigidity reduction best. The most important local feature was subthalamic high-beta power (20-35 Hz). The most important connectivity features were subthalamo-parietal coherence in the very high frequency band (>200 Hz) and subthalamo-parietal coherence in low-gamma band (36-60 Hz). Successful prediction was not due to the model inferring distance to target or symptom severity from neuronal oscillations.

CONCLUSION

This study demonstrates for the first time that neuronal oscillations are predictive of DBS outcome. Coherence between subthalamic and parietal oscillations are particularly informative. These results highlight the clinical relevance of inter-areal synchrony in basal ganglia-cortex loops and might facilitate further improvements of DBS in the future.

Consistent Changes in Cortico-Subthalamic Directed Connectivity Are Associated With the Induction of Parkinsonism in a Chronically Recorded Non-human Primate Model

Parkinson’s disease is a neurological disease with cardinal motor signs including bradykinesia and tremor. Although beta-band hypersynchrony in the cortico-basal ganglia network is thought to contribute to disease manifestation, the resulting effects on network connectivity are unclear. We examined local field potentials from a non-human primate across the naïve, mild, and moderate disease states (model was asymmetric, left-hemispheric dominant) and probed power spectral density as well as cortico-cortical and cortico-subthalamic connectivity using both coherence and Granger causality, which measure undirected and directed effective connectivity, respectively. Our network included the left subthalamic nucleus (L-STN), bilateral primary motor cortices (L-M1, R-M1), and bilateral premotor cortices (L-PMC, R-PMC). Results showed two distinct peaks (Peak A at 5–20 Hz, Peak B at 25–45 Hz) across all analyses. Power and coherence analyses showed widespread increases in power and connectivity in both the Peak A and Peak B bands with disease progression. For Granger causality, increases in Peak B connectivity and decreases in Peak A connectivity were associated with the disease. Induction of mild disease was associated with several changes in connectivity: (1) the cortico-subthalamic connectivity in the descending direction (L-PMC to L-STN) decreased in the Peak A range while the reciprocal, ascending connectivity (L-STN to L-PMC) increased in the Peak B range; this may play a role in generating beta-band hypersynchrony in the cortex, (2) both L-M1 to L-PMC and R-M1 to R-PMC causalities increased, which may either be compensatory or a pathologic effect of disease, and (3) a decrease in connectivity occurred from the R-PMC to R-M1. The only significant change seen between mild and moderate disease was increased right cortical connectivity, which may reflect compensation for the left-hemispheric dominant moderate disease state.