Non-linear Functional Brain Co-activations in Short-Term Memory Distortion Tasks

Anna Ceglarek, Jeremi K. Ochab, Ignacio Cifré, Magdalena Fąfrowicz, Barbara Sikora-Wachowicz, Koryna Lewandowska, Bartosz Bohaterewicz, Tadeusz Marek, Dante R. Chialvo

TL;DR

This study introduces a novel non-linear correlation analysis method to investigate how the time of day affects brain activity during short-term memory tasks, specifically focusing on false memory formation. The findings reveal that the time of day influences neural patterns related to memory encoding and retrieval, with distinct brain regions showing different responses to true vs. false memory probes.

What Problem Does This Paper Address?

Understanding how our brains form memories, especially false ones, is a complex challenge. Previous research has explored the neural basis of true and false recognition, and also noted that our cognitive performance fluctuates throughout the day due to circadian rhythms. However, there's a gap in understanding how these two factors – memory distortion and time of day – interact at the neural level. Specifically, how does the brain's functional connectivity, which underpins memory processes, change depending on whether it's morning or evening, and how does this relate to the formation of false memories? This paper aims to bridge this gap by applying a new analytical technique to fMRI data collected during a memory task at different times of the day.

Methods

The core innovation of this study lies in its analytical approach: a non-linear correlation analysis adapted from resting-state fMRI research for task-based paradigms. Traditionally, functional connectivity is assessed using linear Pearson correlations between brain region time series. This paper, however, utilizes a method that focuses on detecting brief, high-amplitude "events" in the BOLD signal. These events are then correlated with specific stimuli presented during the task. This non-linear filtering approach is designed to enhance the signal-to-noise ratio and capture more dynamic functional relationships.

The study employed a modified Deese-Roediger-McDermott (DRM) paradigm, a well-established method for inducing false memories. Participants performed this task twice: once in the morning and once in the evening. The fMRI data collected during these sessions were preprocessed using standard pipelines (slice-time correction, realignment, normalization, smoothing, etc.). The key analysis involved:

  1. Defining "source" and "target" events: Stimulus timing was used to define "source" events, and BOLD signals from 90 brain regions (AAL atlas) were used for "target" events.
  2. Non-linear correlation: The non-linear correlation between these source and target events was computed.
  3. General Linear Model (GLM): These non-linear correlation measures were then used as predictors in a GLM to analyze their relationship with response types (correct vs. false), task phases (encoding vs. retrieval), time of day (morning vs. evening), and specific brain regions.

Datasets

The study utilized fMRI data collected from 54 healthy young participants (mean age 24.17 ± 3.56 years) who underwent two experimental sessions (morning and evening). No new benchmarks were introduced; instead, the study applied a novel analysis method to existing fMRI task data.

Key Insights and Ideas

  • Time of Day Matters for Memory: The study strongly suggests that the time of day isn't just a minor factor but significantly modulates neural activity related to memory formation and retrieval. This reinforces previous findings but provides a more nuanced neural perspective.
  • False Memories Have Distinct Neural Signatures: The non-linear correlation analysis revealed that different brain regions are differentially involved in processing true (positive probes) versus false (lure probes) memories. This highlights that false memories aren't simply errors but arise from specific neural processes.
  • Hippocampal Role in False Recognition: The hippocampus and parahippocampal regions showed distinct correlation patterns with memory probes. Higher correlations in these areas predicted more "no" responses to positive probes (suggesting a failure to recognize something truly seen) and fewer "no" responses to lure probes (suggesting more false alarms).
  • Supplementary Motor Area's Role: In contrast to the hippocampus, the supplementary motor area showed an opposite trend, with higher correlations predicting more "yes" responses to positive probes and fewer to lure probes, suggesting its involvement in accurate response selection.
  • Methodological Advancement: The study successfully demonstrates the utility of non-linear correlation analysis for characterizing brain responses during cognitive tasks, extending its application beyond resting-state research.

Key Results

  • Diurnal Modulation: Significant interactions were found between time of day (morning/evening) and non-linear correlation measures, indicating that the brain's functional connectivity patterns during memory tasks differ depending on the time of day.
  • Probe Type Differences: The study identified significant differences in correlation patterns between positive and lure probes, particularly in hippocampal regions during retrieval. Specifically, higher correlations in the right hippocampus and parahippocampal gyri predicted more "no" responses to positive probes and fewer "no" responses to lure probes.
  • Task-Specific Effects: The influence of time of day and probe type on neural correlations was more pronounced in the local information processing task compared to the global processing task.
  • Specific ROIs Identified: Several brain regions, including the hippocampus, parahippocampal gyrus, supplementary motor area, olfactory cortex, and medial orbitofrontal cortex, showed significant differential correlations related to memory probes and time of day.

Limitations and Future Work

  • AAL Atlas Limitation: The study acknowledges the limitation of using the Automated Anatomical Labeling (AAL) atlas, which does not allow for the examination of fine-grained structures like the hippocampus. Future work could employ higher-resolution atlases or voxel-wise analyses.
  • Delay Interpretation: The interpretation of correlation findings is contingent on the timing of the BOLD response relative to stimulus presentation. While the method can incorporate delays, this study focused on a proof of concept. Future research should explicitly model these delays for more precise interpretations.
  • Generalizability: While the study provides a proof of concept for the non-linear correlation method in task paradigms, further validation across different tasks and populations is warranted.
  • Synaptic Homeostasis Hypothesis: The authors suggest that diurnal variations in false memory formation might be explained by the synaptic homeostasis hypothesis, which posits lower cortical inhibition immediately after sleep. This is a promising avenue for future investigation.
  • Broader Applications: The effectiveness of the non-linear correlation method in revealing task-specific neural correlates suggests its potential for analyzing other cognitive tasks and understanding various brain functions.

Summary updated

Abstract

Recent works shed light on the neural correlates of true and false recognition and the influence of time of day on cognitive performance. The current study aimed to investigate the modulation of the false memory formation by the time of day using a non-linear correlation analysis originally designed for fMRI resting-state data. Fifty-four young and healthy participants (32 females, mean age: 24.17 ± 3.56 y.o.) performed in MR scanner the modified Deese-Roediger-McDermott paradigm in short-term memory during one session in the morning and another in the evening. Subjects' responses were modeled with a general linear model, which includes as a predictor the non-linear correlations of regional BOLD activity with the stimuli, separately for encoding and retrieval phases. The results show the dependence of the non-linear correlations measures with the time of day and the type of the probe. In addition, the results indicate differences in the correlations measures with hippocampal regions between positive and lure probes. Besides confirming previous results on the influence of time-of-day on cognitive performance, the study demonstrates the effectiveness of the non-linear correlation analysis method for the characterization of fMRI task paradigms.
BibTeX
@inproceedings{ceglarek2021non,
  author = {Anna Ceglarek and Jeremi K. Ochab and Ignacio Cifré and Magdalena Fąfrowicz and Barbara Sikora-Wachowicz and Koryna Lewandowska and Bartosz Bohaterewicz and Tadeusz Marek and Dante R. Chialvo},
  title = {{Non-linear Functional Brain Co-activations in Short-Term Memory Distortion Tasks}},
  booktitle = {Frontiers in Neuroscience},
  year = {2021},
  doi = {10.3389/fnins.2021.778242},
}
Figure