Poster #9 - Do ERP and fMRI signals explain different sources of individual variability in memory during childhood?

• In Event: Poster Session 03

Abstract

Decades of research have explored neural correlates of memory development using EEG and fMRI. However, connections between findings from the two neuroimaging approaches remain unclear. Emerging research in cognitive neuroscience has adopted a latent variable modeling approach to operationalize neural activations as latent constructs (Cooper et al., 2019), which provides flexibility in investigating sources of individual variability and how this variability relates to behavior. Using SEM, this study sought to explore whether event related potentials (ERPs) and fMRI signals obtained from the same children (1 week apart) explain different variability in young children’s behavioral episodic memory performance.

This study used data from a larger research project examining episodic memory development and the brain in our lab (blinded for review). Previous reports have examined developmental changes in episodic memory performance, ERPs, and fMRI separately. This report capitalized on data from prior work including: 1) a latent memory performance measure estimated from four behavioral memory tasks (blinded for review), 2) clustered ERP late slow wave (LSW) amplitude during a source memory encoding paradigm (blinded for review), and 3) task-fMRI activation in the anterior and posterior hippocampus during the same task paradigm with different memory items (blinded for review). A total of 22 4- to 8-year-old children provided usable data for all 3 domains. Analyses were conducted in a SEM framework. Prior to modeling, exploratory factor analysis (EFA) was performed to investigate latent structures of LSW amplitudes and fMRI hippocampal activations.

A latent factor of hippocampal encoding activation was identified using the contrast between subsequent source correct vs subsequent source incorrect memory trial activation in anterior and posterior hippocampus. However, a latent structure was not identified in source-correct v.s. source-incorrect LSW amplitudes (a finding consistent with a similar analysis in a larger sample of 73 children). Thus, we included the latent factor of hippocampal activation, raw scores of clustered LSW amplitudes, and the covariance between domains in one SEM model, which has good convergence (CFI=1.00, RMSEA=.00, 90%CI=[.00, .20]; SRMR=.02).

Results revealed that the latent hippocampal factor was negatively associated with children’s episodic memory performance (β=-.39, p<.05), suggesting that children with smaller differences in activation between trials had better episodic memory. In addition, several clusters of LSW signals (left-frontal: β=.81, p<.01; left-central: β=-.57, p< .01; left-posterior: β=.30, p<.05; medial-central: β=1.08, p<.001, medial posterior: β=-.51, p<.02) were also associated with memory performance. Finally, there was no significant covariance between the latent hippocampal activation and LSW amplitudes, suggesting that these methodologies account for different sources of variance in episodic memory.

Overall, our results provide an initial step in moving beyond examining the complex relation of the brain and cognition using single measures of activation and memory ability. Although informative, these results were unstable, likely due to the small sample size and high collinearity among LSW data. To address this limitation, future analyses will explore resting state EEG and fMRI data with a larger sample from the same study. Additional modeling methods will include two-stage least squares and exploratory SEM.

Authors

• Yuqing Lei, University of Maryland - College Park
  Presenting Author

• Zehua Cui, University of Maryland - College Park

• Kelsey Leigh Canada, Wayne State University

• Tracy Riggins, University of Maryland - College Park