Patients with Parkinson’s Disease (PD) often experience gait impairments, which severely affect their quality of life. Previous studies have suggested that β-frequency (15-30 Hz) oscillatory activity in the basal ganglia may be associated with gait impairments, but the exact dynamics of these oscillations during the gait process remain unclear. Additionally, existing research has found that auditory cues can improve the gait kinematics of PD patients. If we can better understand the neurophysiological mechanisms underlying these cues, it may be possible to treat gait impairments using adaptive deep brain stimulation (ADBS) techniques. Therefore, this study aimed to characterize the dynamic features of oscillatory activity in the subthalamic nucleus (STN) during the gait process and to explore the neurophysiological mechanisms by which auditory cues modulate gait.

Source and Author Information

This paper was written by Chien-Hung Yeh, Yifan Xu, Wenbin Shi, James J. Fitzgerald, Alexander L. Green, Petra Fischer, Huiling Tan, and Ashwini Oswal, who are affiliated with the School of Information and Electronics, Beijing Institute of Technology; the Key Laboratory of Brain Health Intelligent Assessment and Intervention, Ministry of Education, Beijing Institute of Technology; the Nuffield Department of Surgical Sciences, University of Oxford; the Oxford Functional Neurosurgery, University of Oxford; the School of Physiology, Pharmacology and Neuroscience, University of Bristol; the Brain Network Dynamics Unit, MRC, University of Oxford; and the Nuffield Department of Clinical Neurosciences, University of Oxford. The article was published in the journal Brain Stimulation in April 2024 by Elsevier.

Research Procedure

The study included 8 PD patients who underwent STN local field potential (LFP) recordings while performing gait stepping tasks. Hidden Markov Models (HMMs) were used to discover the instantaneous spectral activity states during the gait process. The experiment was divided into three conditions: 1) pre-sound, 2) on-sound, and 3) post-sound. Data processing and analysis were mainly conducted in MATLAB, using Masking Empirical Mode Decomposition (MEMD) and HMM decoding to characterize the short-term spectral features.

Detailed steps:

1. Patient Recruitment and Data Acquisition:

   • 8 PD patients, with a median age of 61.4 years and an average disease duration of 11.3 years.
   • STN-LFPs were recorded using a bipolar configuration, and the electrode positions relative to the STN were confirmed through post-operative CT and pre-operative MRI scans.
   • Experimental conditions: 90-second periods of pre-sound, on-sound, and post-sound.

2. Data Processing:

   • Raw LFP data underwent 50 Hz notch filtering to suppress power line noise, followed by high-pass filtering and downsampling.
   • Spectrogram of the signal was constructed using continuous Morlet wavelet transform.

3. HMM Analysis:

   • TDE-HMM was used to detect instantaneous states in the STN LFP, and correlation with the Hilbert envelope signal was performed to assign spectral states.
   • Time-domain features of the states, including state occupancy and state lifetimes, were extracted.

Unique Methods and Algorithms:

1. Hidden Markov Model (HMM):

   • HMM is an unsupervised machine learning method used to detect instantaneous states in data, representing different instantaneous spectral contents.
   • The study employed TDE-HMM, which uses time-delayed embedding to improve the accuracy of state detection.

2. Masking Empirical Mode Decomposition (MEMD):

   • MEMD is an improved version of EMD, capable of effectively decomposing the signal into intrinsic mode functions (IMFs), reflecting local features at different time scales.

3. Detection of Transient Events:

   • Traditional thresholding methods struggle to detect transient activities across multiple frequency bands simultaneously. This study used HMM to automatically detect burst-like activities in different bands.

Research Results

1. Improved Gait Performance:

   • Data showed that after auditory cues, the variance of stride intervals and step intervals significantly decreased, indicating that auditory cues improved gait performance in PD patients.

2. Modulation of Frequency Band Activities:

   • State occupancy and state lifetimes of the α band significantly increased during and after auditory cues, suggesting that cues may facilitate movement by enhancing α oscillations.
   • State occupancy and lifetimes of the β band, particularly the low β band, significantly decreased, indicating that auditory cues may suppress excessive β oscillatory activity, thus improving motor impairments.
   • No significant changes were observed in the γ band states, possibly due to the unclear role of γ oscillations in auditory cue modulation.

3. Changes in State Transition Probabilities:

   • During auditory cues, the probability of α states transitioning to background activity increased, while the probability of β states transitioning to γ states decreased, suggesting that auditory cues differentially affected the transition mechanisms across frequency bands.

Conclusion

This paper revealed the modulatory effects of auditory cues on the short-term spectral dynamics of STN LFPs in PD patients. These findings contribute to the understanding of the neurophysiological mechanisms by which auditory cues improve gait and provide a theoretical basis for developing ADBS-based treatment strategies.

Research Highlights

1. Innovative Methods: This study is the first to combine MEMD and TDE-HMM, enabling automatic detection of instantaneous oscillatory activities across multiple frequency bands, overcoming limitations of traditional methods.
2. Detailed Data Analysis: Through detailed analysis of band-specific features and transition probabilities, the study revealed the differential modulatory effects of auditory cues on oscillatory activities in different frequency bands.
3. Practical Value: The research results provide new biomarkers for gait impairments in PD patients and may be applied to adaptive deep brain stimulation treatment strategies.