IMU-Enhanced EEG Motion Artifact Removal with Fine-Tuned Large Brain Models

2025-08-31

AI Summary

Motion artifacts severely degrade EEG signal-to-noise ratio, and conventional unimodal denoising methods (e.g., ASR, ICA) fail to effectively model the complex coupling between motion dynamics and neural activity. Method: We propose an IMU-enhanced multimodal denoising framework that integrates inertial measurement unit (IMU) motion signals—introduced for the first time into EEG denoising—with a fine-tuned LaBraM large model (9.2M parameters) and a correlation-aware attention mechanism to achieve cross-modal temporal alignment and channel-wise artifact identification. Contribution/Results: Our method achieves superior performance over ASR-ICA across multi-task and multi-timescale scenarios, requiring only 5.9 hours of training data (0.23% of baseline), demonstrating exceptional robustness and strong few-shot generalization. It establishes a novel paradigm for resource-efficient, real-world BCI deployment.

Technology Category

• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Computer Vision: Multi-modal Vision

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Electroencephalography (EEG) is a non-invasive method for measuring brain activity with high temporal resolution; however, EEG signals often exhibit low signal-to-noise ratios because of contamination from physiological and environmental artifacts. One of the major challenges hindering the real-world deployment of brain-computer interfaces (BCIs) involves the frequent occurrence of motion-related EEG artifacts. Most prior studies on EEG motion artifact removal rely on single-modality approaches, such as Artifact Subspace Reconstruction (ASR) and Independent Component Analysis (ICA), without incorporating simultaneously recorded modalities like inertial measurement units (IMUs), which directly capture the extent and dynamics of motion. This work proposes a fine-tuned large brain model (LaBraM)-based correlation attention mapping method that leverages spatial channel relationships in IMU data to identify motion-related artifacts in EEG signals. The fine-tuned model contains approximately 9.2 million parameters and uses 5.9 hours of EEG and IMU recordings for training, just 0.2346% of the 2500 hours used to train the base model. We compare our results against the established ASR-ICA benchmark across varying time scales and motion activities, showing that incorporating IMU reference signals significantly improves robustness under diverse motion scenarios.

Problem

• Removing motion artifacts from EEG signals using IMU data
• Improving EEG signal quality for brain-computer interface applications
• Leveraging multi-modal data to enhance motion artifact removal

Innovation

• Fine-tuned large brain model for EEG artifact removal
• Leverages IMU spatial channel correlation attention mapping
• Uses minimal training data relative to base model