Uncovering biomarkers during therapeutic neuromodulation with PARRM: Period-based Artifact Reconstruction and Removal Method

Abstract

SUMMARY Advances in therapeutic neuromodulation devices have enabled concurrent stimulation and electrophysiology in the central nervous system. However, stimulation artifacts often obscure the sensed underlying neural activity. Here, we develop a method, termed Period-based Artifact Reconstruction and Removal Method (PARRM), to remove stimulation artifacts from neural recordings by leveraging the exact period of stimulation to construct and subtract a high-fidelity template of the artifact. Benchtop saline experiments, computational simulations, five unique in vivo paradigms across animal and human studies, and an obscured movement biomarker are used for validation. Performance is found to exceed that of state-of-the-art filters in recovering complex signals without introducing contamination. PARRM has several advantages: (1) it is superior in signal recovery; (2) it is easily adaptable to several neurostimulation paradigms; and (3) it has low complexity for future on-device implementation. Real-time artifact removal via PARRM will enable unbiased exploration and detection of neural biomarkers to enhance efficacy of closed-loop therapies.

PELP: Accounting for Missing Data in Neural Time Series by Periodic Estimation of Lost Packets

“…for appropriate choice of the number of harmonics (m) and the parameter vector ( β j : j = 1,‥.,2 m + 1) (Dastin-van Rijn et al, 2021a ). Within a single run, β can be estimated via least-squares using the harmonic regression model y k,r = ƒ β ( t k,r ) + ɛ k,r with homoscedastic noise ɛ k,r .…”

“…The effect of stimulation was simulated by adding the artifact component regressed from recordings on a different day where stimulation was turned on to data without stimulation. A high-pass filter at 1 Hz with a gaussian window was first applied to achieve approximately 40 dB attenuation in the stopband before the period of stimulation (δ) was identified using the period estimation component of PARRM (Dastin-van Rijn et al, 2021a ). A sum of sinusoids ƒ β ( t ) with m harmonics of the period and coefficients β was then fit to the data using linear regression.…”

“…Although δ is known in principle, slight inaccuracies in device system clocks make it important in practice to use data-driven methods to estimate δ. Before using PELP, we estimate δ from combined data across all runs using the multiple-channel period estimation method described in detail in Dastin-van Rijn et al ( 2021a ) and Provenza et al ( 2021 ) where we treat each run as a separate channel, where we use at most the first 10 4 samples from each run, and where we use only the first two stages of the stagewise search for δ before the final optimization. We similarly choose in a preprocessing step using Akaike Information Criterion (AIC; Akaike, 1974 ) for the harmonic regression model (with Gaussian errors) applied to the single longest run.…”

Molecular Mechanisms and Therapeutic Strategies for Levodopa-Induced Dyskinesia in Parkinson’s Disease: A Perspective Through Preclinical and Clinical Evidence

“…An adaptive DBS study was successfully performed using a novel narrowband gamma oscillation as a biomarker for dyskinesia in the motor cortex to modulate the target stimulation delivered ( Swann et al, 2018 ). They also reported 38–45% less energy consumption than classical DBS to maintain the same therapeutic efficacy ( Dastin-van Rijn et al, 2021 ). Recently, more responsive adaptive DBS systems have been developed using accelerometer-based technology, which detects movements by utilizing peripheral kinematic sensors to detect dyskinesia ( Rojas Cabrera et al, 2020 ).…”

Artifact characterization and mitigation techniques during concurrent sensing and stimulation using bidirectional deep brain stimulation platforms

“…Not only do bidirectional systems provide opportunities for biomarker exploration in ecologically valid environments, a prerequisite for adaptive DBS, but they also allow us to better understand the underlying pathophysiology of neurological and psychiatric disorders ( Gregg et al, 2021 ; Johnson et al, 2021 ; Pal Attia et al, 2021 ; Provenza et al, 2021 ). However, the artifacts introduced by high amplitude, high frequency stimulation present a challenge for analysis and interpretation of the relatively lower amplitude neural signals that we aim to measure ( Zhou et al, 2018 ; Dastin-van Rijn et al, 2021 ). Therefore, it is important to identify potential artifact sources introduced by stimulation, developing techniques to mitigate these artifacts during data collection.…”