Beta-gamma phase-amplitude coupling of scalp electroencephalography during walking preparation in Parkinson’s disease differs depending on the freezing of gait

Yuki Kimoto, Naoki Tani, Takuto Emura, Takahiro Matsuhashi, Takuto Yamamoto, Yuya Fujita, Satoru Oshino, Koichi Hosomi, Hui Ming Khoo, Shimpei Miura, Takahiro Fujinaga, Takufumi Yanagisawa, Haruhiko KishimaFrontiers in Human Neuroscience 2024 · DOI

TL;DR

Researchers have identified a specific brain signal, beta-gamma phase-amplitude coupling (PAC) in the sensorimotor cortex, that can predict freezing of gait (FOG) in Parkinson's disease patients during walking preparation. This finding suggests that cortical PAC could serve as a novel biomarker for FOG, potentially paving the way for real-time interventions to prevent falls.

What Problem Does This Paper Address?

Parkinson's disease (PD) is characterized by motor symptoms like akinesia and rigidity, which have been linked to abnormal beta oscillations in the brain's subthalamic nucleus (STN). While these STN oscillations can be used as biomarkers for deep brain stimulation (DBS) systems, a specific biomarker for freezing of gait (FOG) – a debilitating symptom that severely impacts daily living and increases fall risk – remains elusive. FOG is particularly challenging because it's not solely a motor issue; it's influenced by cognitive and visual factors, suggesting complex network involvement. This paper aims to explore whether scalp electroencephalography (EEG) can capture these network dynamics and identify a reliable biomarker for FOG, particularly during the critical phase of preparing to walk.

Methods

The study utilized scalp EEG recorded from 11 Parkinson's disease patients while they prepared for and performed walking tasks. The core analytical technique was phase-amplitude coupling (PAC), specifically focusing on the relationship between beta (13-30 Hz) and gamma (80-120 Hz) frequency bands in the sensorimotor cortex. PAC quantifies how the phase of a slower oscillation modulates the amplitude of a faster oscillation.

Key methodological aspects include:

  • Continuous EEG Monitoring: EEG was recorded not just during walking but also for 20 seconds before the start of walking, capturing the crucial preparatory phase.
  • Modulation Index (MI): The researchers used the modulation index (MI) to quantify PAC, a standardized measure that accounts for potential biases.
  • Time-Series Analysis: PAC was analyzed in 10-second windows across the entire walking preparation and execution period. This allowed for tracking dynamic changes in brain activity.
  • Machine Learning for Prediction: A Support Vector Machine (SVM) was employed to predict the presence of FOG based on PAC data during the walking preparation phase. This involved training the model on 80% of the data and testing it on the remaining 20%.
  • Clinical Assessments: Patients underwent standard clinical assessments, including the Movement Disorder Society-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS) Part 3 and the Berg Balance Scale, to evaluate motor function and gait difficulty. The Postural Instability and Gait Difficulty (PIGD) score was specifically extracted from the MDS-UPDRS.

Datasets

The study used data collected from 11 patients with advanced Parkinson's disease who were either undergoing or scheduled for STN DBS implantation at Osaka University Hospital. The data consisted of scalp EEG recordings during walking tasks, alongside clinical assessments. No new benchmark datasets were introduced; the study relied on data collected specifically for this research.

Key Insights and Ideas

  • FOG's Preparatory Signature: The most significant insight is that the brain's activity before a person with PD starts walking differs depending on whether they will experience FOG. Specifically, a decrease in beta-gamma PAC in the sensorimotor cortex occurs in individuals who don't freeze, but this decrease is absent in those who do freeze.
  • PAC as a Predictive Biomarker: The study demonstrated that PAC patterns during the walking preparation phase could predict the likelihood of FOG with a notable accuracy of 71.2% using machine learning. This suggests that FOG might be predictable based on brain activity even before the gait event occurs.
  • PAC Dynamics During Gait: While PAC decreases before walking in non-FOG individuals, it increases during walking in those who experience FOG. This dynamic shift highlights how FOG might be associated with altered cortical processing during the act of walking itself.
  • Correlation with Gait Severity: The degree of beta-gamma PAC in the sensorimotor cortex after walking commencement was positively correlated with the PIGD score, indicating that as gait symptoms worsen, this specific brain signal becomes more pronounced.

Key Results

  • Predictive Accuracy: A Support Vector Machine model predicted the presence of FOG with 71.2% accuracy using PAC data from the walking preparation period.
  • PAC Changes:
    • In trials without FOG, beta-gamma PAC in the sensorimotor area decreased during gait preparation (p = 0.011).
  • In trials with FOG, this decrease was not observed (p = 0.64).
  • Beta-gamma PAC increased during walking in trials with FOG (p = 0.0042).
  • Correlation: Beta-gamma PAC 20 seconds after the start of walking showed a positive correlation with the PIGD score (correlation coefficient = 0.406, p = 0.032).
  • No Correlation with Other Scores: PAC did not significantly correlate with MDS-UPDRS Part 3 or Berg Balance Scale scores 20 seconds after walking.

Limitations and Future Work

The authors acknowledge several limitations:

  • Small Sample Size: The study involved only 11 patients, all of whom were being evaluated for DBS. This limits the generalizability of the findings, as the patient population might not represent the full spectrum of PD.
  • Artifacts in EEG: Scalp EEG during walking can be susceptible to movement artifacts. While efforts were made to exclude these, some residual artifacts might have influenced the data.
  • Subjectivity of Assessments: FOG and other gait symptoms are inherently subjective. While two specialists evaluated FOG to minimize variation, objective quantitative measures are still lacking.
  • Limited Off-Medication Data: The study included patients both on and off DBS stimulation but excluded those unable to walk, potentially limiting insights into the off-medication state.

Future work suggested by the authors includes:

  • Expanding BCI Applications: The findings could inform the development of brain-computer interfaces (BCIs) that predict FOG in real-time, enabling immediate therapeutic interventions to prevent falls.
  • Investigating Network Dynamics: Further research into the basal ganglia-cortical PAC relationship and coupling between the sensorimotor cortex and other cortical areas could provide deeper insights into the brain networks involved in FOG.
  • Clinical Translation: Exploring how these PAC biomarkers can be integrated into closed-loop DBS systems, which currently focus on akinesia and rigidity, could lead to more comprehensive PD management.

Summary updated

Abstract

Beta-gamma PAC in the sensorimotor area during preparation for walking differs depending on the emergence of FOG. As gait symptoms worsened, beta-gamma PAC in the sensorimotor area during walking gradually increased. Cortical PAC may be a biomarker for FOG in Parkinson's disease and may lead to the development of strategies to prevent falls in the future.
BibTeX
@inproceedings{kimoto2024beta,
  author = {Yuki Kimoto and Naoki Tani and Takuto Emura and Takahiro Matsuhashi and Takuto Yamamoto and Yuya Fujita and Satoru Oshino and Koichi Hosomi and Hui Ming Khoo and Shimpei Miura and Takahiro Fujinaga and Takufumi Yanagisawa and Haruhiko Kishima},
  title = {{Beta-gamma phase-amplitude coupling of scalp electroencephalography during walking preparation in Parkinson’s disease differs depending on the freezing of gait}},
  booktitle = {Frontiers in Human Neuroscience},
  year = {2024},
  doi = {10.3389/fnhum.2024.1495272},
}
Figure