# Deep neurobehavioral phenotyping uncovers neural fingerprints of locomotor deficits in Parkinson’s disease

## TL;DR

This study uses deep neurobehavioral analysis to identify specific brain activity patterns linked to gait issues in Parkinson’s disease, offering potential for new therapies.

## Contribution

The novel approach combines kinematic and cortical data to uncover distinct neural fingerprints of locomotor deficits in Parkinson’s disease.

## Key findings

- Gait, akinesia, and stationary movements occupy distinct regions in low-dimensional embedding space.
- Hjorth complexity and mobility features are modulated at akinesia onset.
- Neural features in Parkinson’s patients partially reflect rodent model findings.

## Abstract

Gait deficits present an unresolved therapeutic challenge in Parkinson’s disease. At the behavioral level, symptoms exhibit heterogeneity, including bradykinesia and hypokinesia during cyclical limb movements, and sudden, involuntary interruptions in the gait sequence, known as freezing of gait. The neural activities driving these various deficits remain largely unknown. Here, we investigated the neural correlates of gait sequence interruptions with a deep phenotyping approach. For this, we transformed kinematic trajectories and cortical oscillations into continuous time series of neurobehavioral features. Next, we combined low-dimensional embedding with supervised classification to identify cortical oscillation features that drive gait deficits. In a rodent Parkinson’s disease model, our approach revealed that gait, akinesia, and stationary movements occupy distinct regions in the low-dimensional embedding space. Among the predominant features separating the states, Hjorth complexity and mobility modulated at akinesia onset. Additionally, we validated our findings in two Parkinson’s patients with freezing of gait, where neural features in STN recordings partially reflected the results in rodents. The presented neurobehavioral phenotyping approach is translational and can easily be generalized to the analysis of other complex movement disorders. Together, our results highlight specific neural features as potential biomarkers that may support the development of adaptive closed-loop algorithms for gait therapy in PD.