1.西安工业大学电子信息工程学院(西安 710021)
2.空军军医大学唐都医院神经外科(西安 710038)
葛顺楠。E-mail:gesn8561@163.com
收稿:2026-01-09,
修回:2026-03-23,
网络首发:2026-08-17,
纸质出版:2026-08-20
移动端阅览
李强,葛顺楠.从脑电图中去除脑深部电刺激伪影研究进展[J].北京生物医学工程,2026,45(4):426-432.
LI Qiang,GE Shunnan.Research progress on removing deep brain stimulation artifacts from electroencephalogram[J].Beijing Biomedical Engineering,2026,45(4):426-432.
李强,葛顺楠.从脑电图中去除脑深部电刺激伪影研究进展[J].北京生物医学工程,2026,45(4):426-432. DOI: 10.3969/j.issn.1002-3208.2026.04.014.
LI Qiang,GE Shunnan.Research progress on removing deep brain stimulation artifacts from electroencephalogram[J].Beijing Biomedical Engineering,2026,45(4):426-432. DOI: 10.3969/j.issn.1002-3208.2026.04.014.
脑深部电刺激(deep brain timulation, DBS)作为帕金森病(Parkinson’s disease, PD)等神经系统疾病的核心神经调控疗法,其疗效机制的阐明及新一代自适应闭环系统(adaptive deep brain stimulation, aDBS)的开发,高度依赖于脑电图(electroencephalogram, EEG)对皮层电生理活动的非入侵性监测。但是,DBS的高频电脉冲(100~250 Hz)及其诱发的直流电(direct current, DC)会在EEG记录中产生严重伪影。这些伪影掩盖了α、θ、β等关键神经信号,极大地限制了EEG在DBS期间的可解释性,并阻碍了相关生物标志物的发展。为了突破这一技术瓶颈,本文系统性回顾并深度对比了去除DBS伪影的前沿方法。不同于早期的数字滤波技术,本文重点综述了基于奇异值分解(singular value decomposition, SVD)的改进型自适应滤波、基于采样与插值的方法、典型相关分析的盲源分离(blind source separation, BSS)技术,以及新兴的混合深度学习(deep learning, DL)数据驱动框架。量化数据分析表明,SVD与自适应滤波的结合可使时域均方误差(mean squared error, MSE)减少95.26%;而在深度学习领域,GCTNet等生成式模型在低信噪比下对常规生理伪影展现了很好的抑制率和信号保真度(RRMSE < 0.2),但受制于DBS开启时纯净脑电基准缺失的物理悖论,此类深层网络目前未被直接用于DBS伪影的端到端重构。当前针对植入式设备的DL模型已转向利用自编码进行伪影分类(F1 > 0.97)以及利用生成式对抗网络(generative adversarial network, GAN)合成仿真数据。本文深度对比指出,单一的算法已无法解决所遇到的问题,构建前端轻量化硬件消隐+后端非线性网络残差精调的多元融合方法,并向自监督学习迁移,是兼顾高信噪比(signal-to-noise ratio, SNR)与毫秒级延迟的必经之路。
As a core neuromodulation therapy for neurological disorders such as Parkinson’s disease (PD), deep brain stimulation (DBS) relies heavily on non-invasive electroencephalography (EEG) monitoring of cortical electrophysiological activity to elucidate its therapeutic mechanisms and facilitate the development of next-generation adaptive closed-loop systems (aDBS). However, the high-frequency electrical pulses (100 to 250 Hz) generated by DBS, along with the induced direct current (DC) shifts, produce severe artifacts within EEG recordings. These artifacts obscure critical neural signals (such as alpha, theta, and beta waves), thereby severely limiting the interpretability of EEG data during DBS sessions and hindering the development of relevant biomarkers. To overcome this technical bottleneck, this paper systematically reviews and comprehensively compares state-of-the-art methods for removing DBS artifacts. Moving beyond early digital filtering techniques, this review focuses on advanced methods such as improved adaptive filtering based on singular value decomposition (SVD), sampling-and-interpolation-based approaches, blind source separation (BSS) techniques utilizing canonical correlation analysis (CCA), and emerging hybrid deep learning (DL) data-driven frameworks. Quantitative analysis indicates that combining SVD with adaptive filtering can reduce the time-domain mean squared error (MSE) by 95.26%. In the realm of deep learning, generative models, such as GCTNet, have demonstrated excellent suppression rates and signal fidelity (RRMSE < 0.2) regarding common physiological artifacts under low signal-to-noise ratio (SNR) conditions. However, constrained by the inherent physical paradox of lacking “clean” EEG baselines while DBS is active, such deep networks have not yet been directly applied to the end-to-end reconstruction of DBS artifacts. Current DL models tailored for implanted devices have shifted toward utilizing autoencoders for artifact classification (achieving F1 scores > 0.97) and employing generative adversarial networks (GANs) to synthesize realistic simulated data. This in-depth comparative analysis concludes that no single algorithm can adequately resolve the multifaceted challenges at hand. Instead, the optimal path forward lies in developing multi-modal fusion strategies—combining lightweight hardware blanking at the front end with nonlinear network-based residual refinement at the back end—and transitioning toward self-supervised learning to simultaneously achieve high signal-to-noise ratios (SNR) and millisecond-level latency.
Lio G, Thobois S, Ballanger B, et al. Removing deep brain stimulation artifacts from the electroencephalogram: issues, recommendations and an open-source toolbox[J]. Clinical Neurophysiology, 2018, 129(10): 2170-2185.
Chen P, Kim T, Dastin-van Rijn E, et al. Periodic artifact removal with applications to deep brain stimulation[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2022, 30: 2692-2699.
Mikroulis A, Lasica A, Filip P, et al. Patient-specific and interpretable deep brain stimulation optimisation using MRI and clinical review data[J]. Frontiers in Neuroscience, 2025, 19: 1661987.
Hariz M, Blomstedt P. Deep brain stimulation for Parkinson’s disease[J]. Journal of Internal Medicine, 2022, 292: 764-778.
Chu C, Zhang Z, Wang J, et al. Deep learning reveals personalized spatial spectral abnormalities of high delta and low alpha bands in EEG of patients with early Parkinson’s disease[J]. Journal of Neural Engineering, 2021, 18(6): 066036.
Acharyya P, Daley KW, Choi JW, et al. Closing the loop in DBS: a data-driven approach[J]. Parkinsonism & Related Disorders, 2025, 134: 107348.
Bédard C, Kröger H, Destexhe A. Does the 1/f frequency scaling of brain signals reflect self-organized critical states?[J]. Physical Review Letters, 2006, 97(11): 118102.
Zhu M, HajiHosseini A, Baumeister TR, et al. Altered EEG alpha and theta oscillations characterize apathy in Parkinson’s disease during incentivized movement[J]. NeuroImage: Clinical, 2019, 23: 101922.
Anis S, Clark PA, Shaffer S, et al. Introducing a workflow algorithm for adaptive DBS programming in Parkinson’s disease[J]. Parkinsonism & Related Disorders, 2026, 144: 108199.
Neumann WJ, Staub-Bartelt F, Horn A, et al. Long term correlation of subthalamic beta band activity with motor impairment in patients with Parkinson’s disease[J]. Clinical Neurophysiology, 2017, 128(11): 2286-2291.
Allen DP, Stegemöller EL, Zadikoff C, et al. Suppression of deep brain stimulation artifacts from the electroencephalogram by frequency-domain Hampel filtering[J]. Clinical Neurophysiology, 2010, 121(8): 1227-1232.
Sun Y, Farzan F, Garcia Dominguez L, et al. A novel method for removal of deep brain stimulation artifact from electroencephalography[J]. Journal of Neuroscience Methods, 2014, 237: 33-40.
Bahador N, Saha J, Rezaei MR, et al. Robust removal of slow artifactual dynamics induced by deep brain stimulation in local field potential recordings using SVD-based adaptive filtering[J]. Bioengineering, 2023, 10: 719.
Liu TC, Chen YC, Chen PL, et al. Removal of electrical stimulus artifact in local field potential recorded from subthalamic nucleus by using manifold denoising[J]. Journal of Neuroscience Methods, 2024, 403: 110038.
Liu TC, Chen PL, Chen YC, et al. Efficient artifact removal for adaptive deep brain stimulation and a temporal event localization analysis[J]. Journal of Neuroscience Methods, 2026, 430: 110708.
Xing C, Cheng X, Feng H, et al. A real-time artifact removal system for closed-loop deep brain stimulation[J]. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 2025, 33: 3237-3245.
Delorme A, Sejnowski T, Makeig S. Enhanced detection of artifacts in EEG data using higher-order statistics and independent component analysis[J]. NeuroImage, 2007, 34(4): 1443-1449.
Onton J, Makeig S. Information-based modeling of event-related brain dynamics[J]. Progress in Brain Research, 2006, 159: 99-120.
Abbasi O, Hirschmann J, Schmitz G, et al. Rejecting deep brain stimulation artefacts from MEG data using ICA and mutual information[J]. Journal of Neuroscience Methods, 2016, 268: 131-141.
Kandemir AL, Litvak V, Florin E. The comparative performance of DBS artefact rejection methods for MEG recordings[J]. NeuroImage, 2020, 219: 117057.
Hofmanis J, Ruiz RAS, Caspary O, et al. Extraction of Deep Brain Stimulation (DBS) source in SEEG using EMD and ICA[C]// 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society. Boston, MA, USA: IEEE Press, 2011: 834-837.
Chuang CH, Chang KY, Huang CS, et al. IC-U-Net: a U-Net-based denoising autoencoder using mixtures of independent components for automatic EEG artifact removal[J]. NeuroImage, 2022, 263: 119586.
Hammer LH, Kochanski RB, Starr PA, et al. Artifact characterization and a multipurpose template-based offline removal solution for a sensing-enabled deep brain stimulation device[J]. Stereotactic and Functional Neurosurgery, 2022, 100(3):168-183.
Mivalt F, Sladky V, Balzekas I, et al. Deep generative networks for algorithm development in implantable neural technology[C]// 2022 IEEE International Conference on Systems, Man, and Cybernetics (SMC). Prague, Czech Republic: IEEE Press, 2022: 1736-1741.
Jiang R, Tong S, Wu J, et al. A novel EEG artifact removal algorithm based on an advanced attention mechanism[J]. Scientific Reports, 2025, 15(1): 19419.
Yin J, Liu A, Li C, et al. A GAN guided parallel CNN and transformer network for EEG denoising[J]. IEEE Journal of Biomedical and Health Informatics, 2023, 29(6): 3930-3941.
Zhang H, Wei C, Zhao M, et al. A novel convolutional neural network model to remove muscle artifacts from EEG[C]// ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). Toronto, ON, Canada: IEEE Press, 2021: 1265-1269.
Kumar G, Zhou Z, Wang Z, et al. Real‐time field‐programmable gate array‐based closed‐loop deep brain stimulation platform targeting cerebellar circuitry rescues motor deficits in a mouse model of cerebellar ataxia[J]. CNS Neuroscience & Therapeutics, 2024, 30(3): e14638.
0
浏览量
0
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621