A cardiac pulse signal affects local field potentials recorded from deep brain stimulation electrodes across clinical targets

Objective Many deep brain stimulation (DBS) systems sense local field potentials (LFPs) for patient monitoring or closed-loop therapy (CL-DBS). LFPs can be impacted by artifacts, including a recently discovered cardiac non-electrocardiographic pulse signal that can be visually masked by commercial device filters. We aimed to establish its prevalence across patient groups and brain areas, and to investigate its spectral impact.

We performed a cross-sectional analysis of LFPs recorded from the cranially mounted Picostim from the pedunculopontine nucleus in multiple systems atrophy patients, periacqueductal gray and sensory thalamus in chronic pain patients, and the centromedian thalamic nucleus (CMT) in paediatric epilepsy patients. For comparison, we analyse externalised recordings from the subthalamic nucleus in Parkinson’s disease patients. The PulsAr algorithm was developed to detect and extract pulsatile signals, and we characterised contamination level and spectral content.

Though not visually obvious in CMT, the pulsatile signal was algorithmically detected in all targets, with 33% of LFPs across targets classed as contaminated. Pulse signal power was similar across targets and may have been masked by higher endogenous activity in CMT. While its dominant frequencies were in the heartbeat range, the signal had spectral content extending up to >10Hz.

A heart pulse signal affects LFP recordings from DBS leads across brain regions and patient groups. While masked by some device filters, spectral content can extend into higher (clinically relevant) frequencies. Researchers and clinicians should exercise caution when sensing lower LFP frequencies, especially for automated control of therapy in CL-DBS. Highlights Historically, electrocardiographic artifacts have been a major source of artifact affecting deep brain stimulation recordings, however pulsatile artifact is less well described. A cardiac pulse signal affects local field potentials recorded from deep brain stimulation electrodes across clinical targets. We introduce an ECG-independent algorithm that detects and extracts this pulsatile signal. The heart pulse signal looks like the intracranial pressure waveform and affects spectral frequencies above the heart rate range up to >10Hz. Clinicians should incorporate screening and procedures to ensure accurate biomarker detection for clinical decision making. Competing Interest Statement KT, RP, and MT have no disclosures to declare. TJD is a founder and chief engineer of Amber Therapeutics Ltd., is the non-executive chairman of Mint Neurotechnologies Ltd, and a non-executive director at Onward Medical N.V. JvR has received speaker fees from Medtronic in 2023 and has acted as a paid consultant for Amber Therapeutics Ltd. in 2025. AG is a founding member of Amber Therapeutics and a consultant with Abbott. WJN received honoraria for consulting from InBrain Neuroelectronics that is a neurotechnology company and honoraria for talks from Medtronic that is a manufacturer of deep brain stimulation devices unrelated to this manuscript. Footnotes katherine.tourigny{at}sjc.ox.ac.uk, rory.piper{at}ucl.ac.uk, martin.tisdall{at}gosh.nhs.uk, julian.neumann{at}charite.de, alex.green{at}nds.ox.ac.uk, timothy.denison{at}eng.ox.ac.uk, joram.vanrheede{at}bndu.ox.ac.uk Author affiliations have been updated. Figure 7A has been edited to display data in box plots. Figure 7B is now part of a new figure, figure 8, which displays high pass filtering of signals at different filter levels. Changes to the results and discussion have been made to incorporate analysis of figure 8. The previous figure 8 has been edited to expand on commentary in the discussion and is now figure 9. Table 3 and the corresponding text has been edited to include re-analysis of data using Kruskal-Wallis tests. Edit to generative AI statement. The major conclusions from the paper have not changed.