Convolutional Monge Mapping between EEG Datasets to Support Independent Component Labeling

2025-09-01

AI Summary

EEG data suffer from device-specific variations (e.g., sensors, amplifiers, filters) and noise, degrading the performance of automatic independent component (IC) classification and undermining artifact removal reliability. To address this, we propose Convolutional Monge Mapping-based Spectral Normalization (CMMN), a novel method that introduces two reference spectrum computation strategies: (i) channel-averaged and L1-normalized barycentric spectra, and (ii) cross-subject nearest-spectrum matching. CMMN further constructs a spatiotemporally separable filter to achieve robust spectral alignment across EEG recordings with varying numbers of channels. By integrating ICA with deep feature alignment principles, CMMN significantly improves classification accuracy for neural ICs—enhancing discrimination between brain sources and artifacts. It effectively mitigates domain shift induced by hardware and acquisition environment heterogeneity. Experimental results demonstrate superior generalizability, establishing CMMN as a new paradigm for robust, automated EEG artifact removal.

Technology Category

• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Learning with Manifolds
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Web Mining and Content Analysis: Web data integration and cleaning
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

EEG recordings contain rich information about neural activity but are subject to artifacts, noise, and superficial differences due to sensors, amplifiers, and filtering. Independent component analysis and automatic labeling of independent components (ICs) enable artifact removal in EEG pipelines. Convolutional Monge Mapping Normalization (CMMN) is a recent tool used to achieve spectral conformity of EEG signals, which was shown to improve deep neural network approaches for sleep staging. Here we propose a novel extension of the CMMN method with two alternative approaches to computing the source reference spectrum the target signals are mapped to: (1) channel-averaged and $l_1$-normalized barycenter, and (2) a subject-to-subject mapping that finds the source subject with the closest spectrum to the target subject. Notably, our extension yields space-time separable filters that can be used to map between datasets with different numbers of EEG channels. We apply these filters in an IC classification task, and show significant improvement in recognizing brain versus non-brain ICs. Clinical relevance - EEG recordings are used in the diagnosis and monitoring of multiple neuropathologies, including epilepsy and psychosis. While EEG analysis can benefit from automating artifact removal through independent component analysis and labeling, differences in recording equipment and context (the presence of noise from electrical wiring and other devices) may impact the performance of machine learning models, but these differences can be minimized by appropriate spectral normalization through filtering.

Problem

• Improving EEG artifact removal via independent component labeling
• Mapping EEG datasets with different channel counts using CMMN
• Enhancing machine learning performance through spectral normalization

Innovation

• Convolutional Monge Mapping Normalization extension
• Two alternative source reference spectrum approaches
• Space-time separable filters for cross-dataset mapping