Back

Characterizing Deep Brain Stimulation Dual Device Beat Frequency Artifacts

Diab, N.; Presbrey, K.; Cernera, S.; Rajesh, S.; Bechtold, R.; Giridharan, N.; Banks, G.; Storch, E. A.; Wang, D. D.; Starr, P. A.; Goodman, W. K.; Herron, J. A.; Sheth, S. A.; Provenza, N. R.

2025-10-13 neurology
10.1101/2025.10.11.25337803 medRxiv
Show abstract

IntroductionRecording-enabled deep brain stimulation devices are used to treat individuals with a variety of neurological disorders, but current systems are only able to connect to a maximum of two brain leads. Off-label implantation of additional leads for unique clinical research questions requires bilateral implantable pulse generators (IPGs), however, slight mismatches in clock rates across IPGs produce mismatched stimulation frequencies. This mismatch creates high amplitude beat frequency artifacts (BFA) that occur at regular intervals and contaminate underlying neural signals, presenting unique challenges for future adaptive DBS (aDBS) algorithm development utilizing dual IPGs. MethodWe quantified BFA intervals in local field potential (LFP) recordings from 26 patients implanted with two IPGs during both continuous (cDBS; n=21) and adaptive (aDBS; n=5) stimulation modes. ResultsBFAs occur in all patients at varying intervals and require both devices to be turned on. We found that BFAs result from mismatched stimulation frequencies, where smaller frequency differences result in longer intervals between BFAs. The range in interval length between artifacts was 112 seconds to 30 minutes across patients. Switching to adaptive DBS (aDBS) decreased this interval to 30 seconds in a single patient. ConclusionIn a dual-device scenario, BFAs should be considered in LFP analysis or future aDBS algorithm design through implementation of mitigation strategies such as onset times and selection of minimally affected recording channels.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.