Microstate-metric test-retest reliability to assess simultaneous EEG-fMRI noise reduction methods

1685 

Abstract Submission 

Toshikazu Kuroda¹, Takeshi Ogawa², Reinmar Kobler³, Mizuki Tsutsumi¹, Tomohiko Kishi¹, Motoaki Kawanabe¹

¹Advanced Telecommunications Research Institute International, Soraku-gun Seika-cho, Kyoto, ²Advanced Telecommunications Research Institute International, Kyoto, Japan, ³ATR, Seika-cho, Soraku-gun, Kyoto

Toshikazu Kuroda  
Advanced Telecommunications Research Institute International
Soraku-gun Seika-cho, Kyoto

Takeshi Ogawa  
Advanced Telecommunications Research Institute International
Kyoto, Japan

Reinmar Kobler  
ATR
Seika-cho, Soraku-gun, Kyoto

Mizuki Tsutsumi  
Advanced Telecommunications Research Institute International
Soraku-gun Seika-cho, Kyoto

Tomohiko Kishi  
Advanced Telecommunications Research Institute International
Soraku-gun Seika-cho, Kyoto

Motoaki Kawanabe  
Advanced Telecommunications Research Institute International
Soraku-gun Seika-cho, Kyoto

Electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) are two major noninvasive approaches for measuring brain activity. Their simultaneous recording has an advantage in examining how EEG is associated with fMRI. Although promising, the simultaneous recording suffers from serious contamination of noise in EEG. Moreover, it remains difficult to evaluate noise reductions due to the absence of ground truth in EEG. One remedy is the use of EEG recorded outside MR scanners as a reference (van der Meer et al., 2016). This approach implicitly assumes high test-retest reliability, however: EEG should be similar between inside and outside the scanner if noise reductions are sufficient. EEG microstates could be a useful alternative for the evaluation of noise reductions in the context of simultaneous recording with fMRI. Briefly, the microstates are extracted by clustering EEG spatial maps, typically resulting in four to five templates (Michel et al., 2018). Backfitting the templates to raw EEG provides several metrics such as duration, occurrence, and coverage, and transition probability of each microstate. Moreover, the test-retest reliability of the first three metrics generally is high at least for EEG recorded outside the MR scanner (Kleinert et al., 2023).

Here we introduce a new evaluation method for noise reduction in the context of EEG-fMRI simultaneous recording taking advantage of EEG microstate metrics. We assumed that, if EEG recorded outside an MR scanner on two different days has high test-retest reliability in terms of microstate metrics, then it would be reasonable to consider that the reliability should also be high for EEG recorded inside and outside the scanner on the same day. This led us to the notion that the scanning noise should be reduced in a way that the reliability would increase. Taking this approach, we recorded EEG during an 8-min resting state both inside and outside the scanner on two separate days per participant (n=25). For EEG recorded outside the scanner, artifacts were reduced in a similar way to the previous study (Kleinert et al., 2023). EEG recorded inside the scanner was also processed in the same way but with additional steps for reducing scanning-specific artifacts. In particular, residual ballistocardiogram (BCG) artifacts were reduced with an extension of optimal basis set (OBS; de Cheveigné et al., 2014; Niazy et al., 2005) in which a threshold was set for the ratio between EEG power time-locked to heartbeats and the power during randomly selected periods. Using intraclass correlation coefficient (ICC) as an index, test-retest reliability of microstate metrics was assessed 1) between a pair of recordings on Days 1 and 2 outside the scanner, 2) between a pair of recordings on the same day inside and outside the scanner, 3) and between a pair of recordings on Days 1 and 2 inside the scanner.

We obtained moderately high test-retest reliability (ICCs > .50 on average) for duration, occurrence, and coverage in all the three cases (see Figure 1). Moreover, the ICCs were higher with than without the new method for reducing residual BCG.

We demonstrated that scanning noise can be reduced to a level that EEG microstate metrics reached moderately high test-retest reliability between a pair of EEG recorded on the same day inside and outside an MR scanner and also between a pair of EEG recorded on different days inside the scanner. Still lower reliability for EEG recorded inside than outside the scanner suggest a room for further improvement in noise reduction.

EEG/MEG Modeling and Analysis

Exploratory Modeling and Artifact Removal ¹

Methods Development ²

Task-Independent and Resting-State Analysis

EEG

Data analysis

Electroencephaolography (EEG)

FUNCTIONAL MRI