Stimulus-Informed Generalized Canonical Correlation Analysis of Stimulus-Following Brain Responses

2025-05-03 0 0 335.58KB 5 页 10玖币

侵权投诉

Stimulus-Informed Generalized Canonical

Correlation Analysis of Stimulus-Following Brain

Responses

Simon Geirnaert1,2,3, Tom Francart2,3, and Alexander Bertrand1,3

1KU Leuven, Department of Electrical Engineering (ESAT),

STADIUS Center for Dynamical Systems, Signal Processing and Data Analytics, Leuven, Belgium

2KU Leuven, Department of Neurosciences, ExpORL, Leuven, Belgium

3KU Leuven Institute for Artiﬁcial Intelligence (Leuven.AI), Leuven, Belgium

{simon.geirnaert,tom.francart,alexander.bertrand}@kuleuven.be

Abstract—In brain-computer interface or neuroscience appli-

cations, generalized canonical correlation analysis (GCCA) is

often used to extract correlated signal components in the neural

activity of different subjects attending to the same stimulus.

This allows quantifying the so-called inter-subject correlation

or boosting the signal-to-noise ratio of the stimulus-following

brain responses with respect to other (non-)neural activity. GCCA

is, however, stimulus-unaware: it does not take the stimulus

information into account and does therefore not cope well

with lower amounts of data or smaller groups of subjects. We

propose a novel stimulus-informed GCCA algorithm based on

the MAXVAR-GCCA framework. We show the superiority of the

proposed stimulus-informed GCCA method based on the inter-

subject correlation between electroencephalography responses

of a group of subjects listening to the same speech stimulus,

especially for lower amounts of data or smaller groups of subjects.

Index Terms—generalized canonical correlation analysis, elec-

troencephalography, stimulus-following neural response

I. INTRODUCTION

Traditionally, several neuroscience applications and brain-

computer interface technologies heavily rely on active partici-

pation of the user following speciﬁc instructions. Moreover,

synthetic and controlled stimuli such as ﬂickering visual

patterns or beep tones are used, as they evoke more controlled

brain responses. Lastly, they rely on multi-trial experiments,

where the same controlled stimulus is repeatedly presented to

allow enhancing the signal-to-noise ratio (SNR) by averaging

the responses across multiple trials [1]. In the past few

years, a shift has been occurring towards passive, single-trial

experiments using natural sensory stimuli, such as speech

or video footage [2]–[8]. This shift opens doors to novel

applications, for example, in attention tracking in hearing de-

vices [2], [3], the (online) classroom [4], neuromarketing [5],

This research is funded by a PDM mandate from KU Leuven (for S.

Geirnaert, No PDMT1/22/009), FWO project nr. G081722N, the European

Research Council (ERC) under the European Union’s Horizon 2020 research

and innovation programme (grant agreement No 802895), and the Flemish

Government (AI Research Program). The scientiﬁc responsibility is assumed

by its authors.

[6], or virtual reality environments [9]. Given that we are

interested in stimulus-following brain responses, we focus

on electroencephalography (EEG) or magnetoencephalography

(MEG), which have an excellent temporal resolution [1].

This shift brings along several signal processing-related

challenges, such as the strong subject-speciﬁcity arising from

using uncontrolled and natural stimuli that evoke highly vari-

able responses, and especially the low SNR of the stimulus-

following EEG responses buried under interfering non-neural

and neural activity that is not time-locked to the stimulus.

In this single-trial setting, the low SNR can not be dealt

with anymore by averaging multiple trials, such that data-

driven ﬁltering methods are required to enhance the stimulus-

following neural response and suppress all other noise sources.

In this paper, we focus on generalized canonical corre-

lation analysis (GCCA), which allows extracting correlated

components across multiple (EEG) recordings, for example

from different subjects. GCCA can then be used not only to

enhance the SNR but also to quantify inter-subject correlation

(ISC) [4]–[7], for dimensionality reduction, or to summarize a

set of EEG recordings [10]. An overview of GCCA for brain

data analysis can be found in de Cheveign´

e et al. [10].

We focus on the problem of extracting the stimulus-

following neural response from a set of EEG recordings from

multiple subjects attending to the same stimulus. One of the

strengths of GCCA here is that it is stimulus-unaware: it

makes no assumptions about the stimulus (representation) and

also works when the stimulus is unavailable. However, this

strength can be a weakness at the same time. As explained

before, one of the main challenges is the notoriously low SNR

of the stimulus-following neural response. Maximally exploit-

ing all available side information, including the stimulus, is

paramount to optimally boost the SNR, even more so when

only a few subjects or small amount of data are available.

The latter regularly occurs, for example, in a time-adaptive

context for online processing [3]. Therefore, we modify the

GCCA formulation to include the stimulus as side-information,

leading to stimulus-informed GCCA (SI-GCCA).

arXiv:2210.13297v3 [eess.SP] 16 Feb 2023

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Stimulus-InformedGeneralizedCanonicalCorrelationAnalysisofStimulus-FollowingBrainResponsesSimonGeirnaert1,2,3,TomFrancart2,3,andAlexanderBertrand1,31KULeuven,DepartmentofElectricalEngineering(ESAT),STADIUSCenterforDynamicalSystems,SignalProcessingandDataAnalytics,Leuven,Belgium2KULeuven,Departmentof...

展开>> 收起<<

Stimulus-Informed Generalized Canonical Correlation Analysis of Stimulus-Following Brain Responses.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Stimulus-Informed Generalized Canonical Correlation Analysis of Stimulus-Following Brain Responses

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: