eeg preprocessing and quality control

This study addresses the challenge posed by strong artifacts in transcranial magnetic stimulation (TMS)-evoked electroencephalography (EEG) signals, which hinder their application in closed-loop neuromodulation and brain–computer interfaces. The work presents the first standardized benchmark dataset for TMS-EEG denoising and systematically evaluates two mainstream source-domain denoising pipelines under conditions lacking ground-truth physiological signals, assessing both artifact suppression efficacy and preservation of genuine TMS-evoked responses. The proposed preprocessing framework demonstrates robust performance, significantly enhancing signal quality and establishing a unified benchmark for algorithm development. This advancement facilitates more reliable use of TMS-EEG in neuroscience research, clinical settings, and embedded brain–computer interface systems.

EEG/LFP signals are often severely contaminated by artifacts and noise, and conventional preprocessing relies heavily on manual intervention, compromising reproducibility. To address this, we propose a fully automated, unsupervised five-step pipeline: integrated bandpass and notch filtering, automatic bad-channel detection, ICA decomposition, and clustering-based automatic component classification. Our key contribution lies in the deep integration of ICA with unsupervised machine learning to achieve high-accuracy artifact identification and removal while preserving task-relevant neural information—thereby substantially reducing human bias. Evaluated across multiple heterogeneous datasets, our method improves classification accuracy of downstream logistic regression models—from 74% with conventionally preprocessed data to over 97%—demonstrating marked gains in performance and reliability for brain–computer interface and neural decoding applications.

To address the performance degradation of electromyographic (EMG) artifact removal in electroencephalography (EEG) under high-noise conditions, this paper proposes a novel denoising framework based on a statistical Mixture of Experts (MoE). The method employs a CNN-RNN hybrid architecture for expert networks and introduces three key innovations: (1) the first quantitative taxonomy of EMG artifact subtypes; (2) localized expert models specifically tailored to distinct signal-to-noise ratio (SNR) regimes; and (3) a correlation-driven objective function coupled with a rescaling strategy to accelerate convergence and improve the denoising lower bound under severe noise. Evaluated on the EEGdenoiseNet benchmark, the framework achieves state-of-the-art (SOTA) overall performance while significantly enhancing the lower-bound PSNR and SSIM—particularly in high-noise scenarios—thereby improving the robustness and control accuracy of neural interfaces.

Medical time series such as electrocardiograms (ECGs) and photoplethysmograms (PPGs) are frequently affected by measurement artifacts due to challenging acquisition environments, such as in ambulances and during routine daily activities. Since automated algorithms for analyzing such signals increasingly inform clinically relevant decisions, identifying signal segments on which these algorithms may produce unreliable outputs is of critical importance. Signal quality indices (SQIs) are commonly used for this purpose. However, most existing SQIs are task agnostic and do not account for the specific algorithm and performance metric used downstream. In this work, we formalize signal quality as a task- and metric-dependent concept and propose a perturbation-based SQI (pSQI) that aims to detect an algorithm's performance degradation on an input signal with respect to a metric. The pSQI is defined as the worst-case value of the performance metric under an additive, colored Gaussian noise perturbation with a lower-bounded signal-to-noise ratio. We introduce formal requirements for task- and metric-specific SQIs, including monotonicity of the metric in expectation and maximal separation under thresholding. Experiments on R-peak detection and atrial fibrillation classification benchmarks demonstrate that the proposed pSQI consistently outperforms existing feature- and deep learning-based SQIs in identifying unreliable inputs without requiring training.

In highly regulated domains such as finance, data quality control (QC) is often fragmented into isolated preprocessing steps, undermining end-to-end trustworthy AI pipelines. To address this, we propose the first AI-driven DataOps framework that embeds QC as a system-level core component. Our framework deeply integrates rule-based engines, statistical analysis, and custom AI-powered anomaly detection across the entire data lifecycle—from ingestion and transformation to model deployment—enabling dynamic remediation, policy-configurable workflows, and end-to-end auditability. Technically, it unifies data profiling, stream processing, cloud-native storage interfaces, and a proprietary AI detection module. Evaluated in a real-world financial production environment, the framework achieves significantly improved anomaly recall, reduces manual intervention by 42%, and ensures audit completeness and full data traceability under high-throughput conditions, fully satisfying regulatory compliance requirements.

This study addresses the poor interpretability of existing automated models for detecting epileptiform discharges in scalp electroencephalography (EEG) by proposing the first large language model (LLM)-agent-driven, closed-loop program synthesis framework. This framework automatically generates, executes, and optimizes interpretable, deterministic signal processing feature modules. Integrating gradient-boosted tree classifiers with a structured performance feedback mechanism, the approach enables auditable EEG spike detection and incorporates an artifact-aware feature generation strategy to enhance robustness. Evaluated via five-fold cross-validation on the VEPISET dataset, the model achieves an AUC of 0.935, balanced accuracy of 0.699, and an F1 score of 0.557; at a sensitivity of 0.80, it attains a mean precision of 0.470 and specificity of 0.900.

This study addresses the lack of standardized task specifications in existing public electroencephalography (EEG) datasets, which hinders the conversion of heterogeneous data into reusable and auditable benchmark units. To resolve this, the authors propose a framework that integrates a structured task specification language with a shared rulebook, aligning task documentation with executable task kernels to enable standardized EEG task definitions. The work introduces two novel mechanisms—NeuroDoc for rule-guided documentation drafting and NeuroAudit for community review and version control—thereby establishing the first executable and auditable EEG benchmarking system. The released benchmark corpus comprises 53 review criteria and 245 task definitions, demonstrating strong reusability and executability across four foundational EEG models.

This study addresses the lack of integrated workflows in existing EEG toolkits that seamlessly bridge offline analysis and real-time, quality-gated cognitive load assessment, which typically requires manual integration of signal acquisition, quality control, and feature extraction. The authors propose the first end-to-end, quality-gated real-time EEG cognitive load analysis pipeline, incorporating preprocessing, dynamic Alpha-band analysis, task-versus-rest comparisons, a low-latency web API, and an interpretability layer powered by a local large language model (LLM), complemented by an interactive dashboard. Evaluated on a small dataset of 18 recordings, the system successfully replicated 10 within-subject contrasts, with 7 showing expected posterior Alpha suppression during task conditions, thereby demonstrating both reproducibility and real-time performance while effectively bridging the gap between offline research and online applications.

This study addresses the prevailing tendency in EEG denoising research to prioritize model scale while overlooking the relationship between model capacity and downstream task utility, as well as the limited relevance of reconstruction metrics to actual neural signal fidelity. Under fixed architecture, loss function, and training protocol, the authors construct an ultra-compact depthwise-separable convolutional U-Net with parameter counts ranging from 1.05K to 40.26K by varying channel width. Systematic evaluation reveals that standard EEG denoising benchmarks saturate at merely 3–6.5K parameters, and crucially, reconstruction quality exhibits a significant disconnect from downstream BCI performance—optimizing reconstruction metrics actually degrades CSP+LDA classification accuracy (0.547 vs. 0.612). The proposed model, requiring only 33–46KB of memory, is highly suitable for edge deployment.