TL;DR
This paper tackles the challenge of pinpointing the optimal stimulation contact in Parkinson's disease Deep Brain Stimulation (DBS) by developing and validating methods to estimate "pseudo-monopolar" beta power from bipolar recordings. They found that a distance-weighted approach best reconstructs the true beta distribution, and that maximal beta power is more stable in ring than directional contacts, offering a more reliable guide for programming.
What Problem Does This Paper Address?
Deep Brain Stimulation (DBS) is a crucial treatment for Parkinson's Disease (PD), but programming the stimulation parameters can be complex. While subthalamic nucleus (STN) beta activity is a known biomarker for PD motor symptoms and can guide programming, current DBS systems record local field potentials (LFPs) using bipolar contacts. This means the recorded signal is a difference between two contacts, making it difficult to attribute the beta activity to a specific contact. This ambiguity hinders precise programming, especially with advanced directional leads that offer more nuanced stimulation. This paper aims to bridge this gap by developing and validating methods to estimate the beta power at individual contacts (pseudo-monopolar) from these bipolar recordings, thereby improving the accuracy and automation of DBS programming.
Methods
The researchers validated three methods for estimating pseudo-monopolar beta power from bipolar LFP recordings:
- Euclidean Method: This novel approach (developed by the authors) weights spectral power based on the inverse squared Euclidean distance between the averaged recording site and the contact of interest. It utilizes all available bipolar LFPs from segmented contacts.
- Strelow et al. Method: This method uses two horizontal segmented channels and weights their spectrograms by the absolute distance between the contact of interest and the other bipolar recording contact.
- Busch et al. Method: This approach combines the power spectra from a selected ring channel with the percentage spectrum of a vertical segmented channel that includes the contact of interest.
They compared these methods against a "ground truth" derived from externalized recordings (EXT-1) and also evaluated the temporal stability of beta power in short (20-second) windows.
Datasets
The study utilized LFP data from 39 Parkinson's disease patients implanted with bilateral subthalamic directional electrodes and Percept™ implantable pulse generators (IPGs). Data was collected at two time points:
- Externalized Recordings (EXT): Shortly after lead implantation, allowing for longer recording durations (2 minutes). This was used to establish a "ground truth" for monopolar estimations (EXT-1) and to simulate bipolar IPG recordings (EXT-2).
- IPG Recordings: 12 months post-implantation, using the standard bipolar recording configuration of the IPG.
Key Insights and Ideas
- Distance Matters: The Euclidean and Strelow methods, which incorporate distance weighting, showed better agreement with the "real" beta distribution than the Busch method. This highlights the importance of spatial information when reconstructing contact-specific power from bipolar signals.
- Ring vs. Directional Stability: Maximal beta power was significantly more stable across short (20-second) windows in ring contacts compared to segmental (directional) contacts. This suggests that ring contacts might be more reliable for short-duration, automated programming.
- Clinical Relevance: The estimated maximal beta power contacts showed a meaningful alignment with clinically active stimulation contacts one year after surgery, suggesting these methods can indeed guide programming towards effective stimulation sites.
- Short Windows Can Work (with caveats): While short windows (20s) can be sufficient, especially for ring contacts, multiple repetitions of these short recordings are recommended to increase certainty in identifying the maximal beta channel, particularly for segmental contacts.
Key Results
- The Euclidean and Strelow methods showed significantly higher correlation coefficients with the "real" EXT-1 beta distribution compared to the Busch et al. method.
- The Euclidean method correctly identified the #1 beta contact in 54% of hemispheres and both top beta contacts in 32% of hemispheres when compared to EXT-1.
- Maximal beta contacts derived from IPG recordings aligned with clinically active stimulation contacts in approximately 37-67% of hemispheres.
- For ring channels, using three 20-second windows provided about 90% certainty in correctly identifying the #1 beta channel. For segmental channels, achieving 60% certainty required at least one window, with higher numbers of windows improving accuracy.
- The coefficient of variance in beta power ranks was significantly lower for 20-second windows compared to 5-second windows, indicating improved temporal stability.
Limitations and Future Work
- Microlesion Effects: The study acknowledges that early postoperative recordings (EXT) might be influenced by microlesion effects, which could affect beta suppression.
- Spatial Ambiguity: Reconstructing contact-wise power distributions from bipolar signals inherently involves ambiguity, leading to moderate correlations between estimated and true beta distributions. This means pseudo-monopolar estimates should be viewed as probabilistic guidance rather than definitive predictors.
- Directional Stimulation Underutilization: The study noted that directional stimulation was used in only 20% of hemispheres, which might have obscured potential directional fluctuations. This further emphasizes the need for automated programming tools to make directional stimulation more accessible.
- Validation of Repeated Short Recordings: The study suggests repeating short IPG recordings to improve accuracy, but this strategy needs to be validated in future studies.
- Future Directions: The authors suggest that further research is needed to validate the proposed strategy of repeated short recordings and to explore how these findings can be integrated into automated DBS programming algorithms, especially for directional leads.
Summary updated