Crunching the numbers. Responsibly.
Accurate subthalamic beta activity could guide deep brain stimulation programming in Parkinson's disease, but bipolar recordings complicate contact selection. In 39 patients, we validated three methods to estimate pseudo-monopolar beta power. Distance-weighted methods (Euclidean, Strelow) agreed consistently with the externalized "ground-truth" beta distribution. Maximal beta power across 20s windows was more stable in ring than in directional channels. Beta-contacts from all methods aligned with clinically active stimulation contacts one year after surgery.
Distance-weighted methods (Euclidean, Strelow) agreed consistently with the externalized “ground-truth” beta distribution.
moderate variable correlations, non-significant method differences, and inconsistent chance-test p-values limit the comparative claim
Maximal beta power across 20s windows was more stable in ring than in directional channels.
within-recording rank variability supports the direction, though overlapping externalized windows limit deployment generality
Beta-contacts from all methods aligned with clinically active stimulation contacts one year after surgery.
the single top contact exceeded chance only for Euclidean, while broader two-contact overlap supported all methods
Derived from the full evaluation — not a separate score.
Strengths
All three reconstruction methods are applied to the same EXT-2 recordings and compared against the same EXT-1 pseudo-reference, allowing a controlled head-to-head assessment of contact rankings.
↳ Methods, EXT LFPs and Feature extraction; Results, Fig. 2a
The analysis extends contact-ranking work to segmented contacts, where bipolar sensing makes assignment of beta power to individual contacts ambiguous and programming is operationally burdensome.
↳ Introduction; Discussion, final paragraph
The limitations section explicitly characterizes moderate, variable correlations and 50–60% maximal-contact agreement as probabilistic guidance rather than definitive prediction.
↳ Limitations, paragraph 1
Limitations
Several binomial p-values in the EXT-1 comparison do not reconcile with the chance probabilities specified in Methods. Correcting them could change which methods exceed chance and weaken the claimed comparative ranking.
↳ Methods, Statistics; Results, pseudo-monopolar comparison; Fig. 2a
The proposed repeated-acquisition strategy is derived from overlapping 20-second windows within a single two-minute externalized recording that also defines the maximal channel. It has not been validated using prospective repeated IPG acquisitions.
↳ Methods, EXT LFPs and Statistics; Results, Fig. 2d; Limitations, paragraph 2
Moderate and highly variable correlations, non-significant between-method differences, and method-specific clinical-overlap results do not establish consistent agreement, universal alignment, or overall superiority as broadly stated.
↳ Abstract; Results, Fig. 2a and clinical relevance paragraph; concluding Results paragraph
The paper addresses a concrete programming problem and provides a useful controlled comparison of three reconstruction methods on shared recordings. Its methodology is assessable, but the EXT-1 and EXT-2 configurations share source recordings, and the repeated-recording protocol is simulated from overlapping epochs rather than independent acquisitions. The reported chance-test p-values in Fig. 2a require correction because they do not reconcile with the chance levels specified in Methods and may alter the comparative interpretation. The limitations section is substantive, but the abstract and concluding recommendations remain stronger than the moderate, variable agreement and preliminary deployment evidence support.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.1
Confidence mediumThe study extends pseudo-monopolar reconstruction comparisons to directional contacts and links them to recording stability and chronic stimulation choices. The advance remains bounded because the methods are pre-existing, the cohort partly overlaps prior work, and comparative superiority is not statistically established.
While that work focused on vertical contact levels, our study extends these efforts to the segmented contacts.
The recording configurations, reconstruction algorithms, exclusions, ranking outcomes, and chance models are described in assessable detail. Confidence is reduced by the imperfect shared-recording reference, overlapping-window simulation, limited precision, unspecified independence of clinical programming, and p-values that do not reconcile with stated chance levels.
The chance levels were 60% for at least one contact, 16.7% for the #1 contact, and 6.7% for both contacts.
The progression from reconstruction validation to temporal stability and clinical overlap is easy to follow. Central phrases such as “agreed consistently,” “all methods aligned,” and “overall superior” imply stronger or broader support than the moderate correlations and outcome-specific significance warrant.
however, no statistically significant differences were observed between the methods.
The discussion engages prior reconstruction and contact-prediction work and explicitly frames 50–60% agreement as probabilistic guidance. However, the recommendation of a repeated-recording protocol is stronger than the limitations section's acknowledgement that the simulated strategy still requires validation.
The suggested strategy of repeating multiple 20-s IPG recordings needs to be validated in future studies.
Concerns4 of 4 checks
Several reported binomial p-values do not reconcile with the stated chance probabilities, and the discrepancy bears directly on the comparative interpretation of the reconstruction methods. Central language also presents moderate, variable agreement and non-significant method differences as consistent or superior performance.
Ethics approval, Declaration of Helsinki adherence, and written informed consent are declared. No text-grounded research-conduct concern was identified; unavailable conflict, registration, data, and code information remains unverified rather than adverse.
Flags: 1 declared / 5 total
24 of 24 checkable references verified
24 references in manuscript
No retraction notice found in Retraction Watch.
Sources: Retraction Watch ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The analysis includes recordings from the implanted Percept device and compares beta-ranked contacts with 12-month stimulation configurations. Readiness remains preliminary because the repeat-recording recommendation is simulated from one externalized recording rather than prospectively tested as repeated IPG acquisitions.
thereby simulating repeated short acquisitions.
AI-generated, human-governed. Something look off? Contact us to request a review.