Trauma-related intrusive memories (TR-IMs) are a hallmark symptom of post-traumatic stress disorder (PTSD) and other trauma-related disorders, significantly impacting the lives of affected individuals. They are characterized by involuntary, spontaneous, vivid sensory experiences that often lack contextual details, creating a sense of reliving the traumatic event in the “here-and-now.” Existing models, such as the warning signal hypothesis and the dual representation theory, attempt to explain these phenomenological properties. The warning signal hypothesis emphasizes the exaggerated emotional intensity of TR-IMs as a learned cue to acquired threat, while the dual representation theory suggests that TR-IMs lack the necessary contextual details to bind exaggerated sensory details in time and space, leading to vivid sensory experiences and re-experiencing. Neurobiological accounts of TR-IMs typically focus on the interaction between the hippocampus (HPC) and other brain regions. The dual representation theory, for instance, implicates a hyperactive sensory cortex and salience network (SN) in the processing of exaggerated sensory details, while suggesting that hippocampal dysfunction may contribute to the deficient contextual details. Existing neuroimaging research often focuses on the frequency and intensity of TR-IMs but lacks a detailed understanding of the neural correlates of their unique phenomenological properties, particularly concerning the dynamic interactions between the anterior and posterior parts of the HPC and distributed cortical systems. This study aims to identify spatiotemporal patterns of intrinsic HPC-cortical co-activation associated with different phenomenological properties of TR-IMs.

The literature on the neurobiology of intrusive memories highlights the role of several brain regions, including the amygdala, hippocampus, and prefrontal cortex. Studies have shown that the amygdala is involved in the emotional processing of traumatic memories, the hippocampus in contextual binding and memory consolidation, and the prefrontal cortex in cognitive control and emotional regulation. However, much of the previous research on TR-IMs primarily focuses on their frequency and intensity, often utilizing retrospective self-report measures which are susceptible to biases. There is limited understanding of the neural correlates of the unique phenomenological properties of TR-IMs, especially the distinct sensory and emotional aspects and the experience of reliving. Moreover, the functional heterogeneity of hippocampal subregions and the dynamic nature of their interactions with cortical systems are often overlooked in previous studies. Canonical models of PTSD emphasize the role of dysfunctional amygdala-hippocampal-medial prefrontal cortex circuits and disruptions in large-scale neural networks, including elevated activity in the salience network and sensory systems, and decreased connectivity in the default mode network and hippocampus. These disruptions are believed to contribute to the pathophysiology of PTSD and the emergence of intrusive memories. Extant research on episodic memory emphasizes the dynamic interplay between the hippocampus and other brain regions, including the default mode network, salience network, and sensory cortices. Different aspects of episodic memory are supported by distinct patterns of connectivity with the anterior and posterior segments of the hippocampus, where the posterior hippocampus is associated with sensory-perceptual aspects of memory and the anterior hippocampus with cognitive-affective features. This functional heterogeneity provides a framework for investigating the neural substrates of the distinct phenomenological properties of TR-IMs.

This study employed a multi-method approach combining ecological momentary assessments (EMAs) with resting-state functional magnetic resonance imaging (rs-fMRI) to investigate the neural correlates of trauma-related intrusive memories (TR-IMs). Ninety-nine trauma-exposed adults were recruited, with 84 having usable rs-fMRI data. Participants completed two weeks of daily EMAs using a smartphone app, rating the phenomenological properties of their TR-IMs (vividness, visual detail, reliving, emotional intensity, fragmentation, intrusiveness) on a 0-4 Likert scale. At Visit 2, participants underwent a clinical interview (CAPS-5), self-report questionnaires (AMQ), and a 13-minute eyes-open rs-fMRI scan. The rs-fMRI data were preprocessed using fMRIPrep and CONN toolboxes. Co-activation pattern (CAP) analysis was used to assess the spatiotemporal dynamics of anterior/posterior hippocampal (a/pHPC)-cortical networks. Anterior and posterior hippocampal seeds were defined based on prior literature, and a union seed-based approach was used to identify volumes exceeding an activation threshold. Spatial patterns of co-active regions were then clustered into CAPs using k-means clustering, with consensus clustering determining the optimal number of clusters (k=4). CAP metrics, including count (total number of supra-threshold volumes) and persistence (probability of remaining in a given CAP across consecutive volumes), were calculated for each participant. Partial correlations, controlling for age and sex, were used to examine the cross-sectional associations between CAP metrics and TR-IM properties. Linear regression models were then used to assess the specificity of these associations. Repeated measures analysis using linear mixed models (LMMs) was conducted to examine the longitudinal associations. Finally, correlation analyses were performed between CAP metrics and conventional clinical measures (TR-IM frequency, retrospective reports of TR-IM properties, CAPS-5 symptom severity) to compare ecological assessments with traditional methods. Additional analyses examined the moderating effect of PTSD diagnosis.

The study identified four distinct co-activation patterns (CAPs) reflecting the dynamic interplay between the HPC and various brain networks. CAP1, predominantly associated with the aHPC, involved activation of the DMN and deactivation of attentional networks. CAP2, largely associated with the pHPC, showed activation of the visual cortex. CAP3 showed deactivation of the visual cortex, and CAP4 showed activation of the visual cortex, sensorimotor areas, and the salience/ventral attention network (SN/VAN). The frequency of CAPs was associated with several TR-IM properties. Emotional intensity was negatively correlated with the frequency of CAP1 (aHPC-DMN), indicating that lower emotional intensity was associated with a more frequent occurrence of this resting-state pattern. Conversely, visual properties were positively correlated with the frequency of CAP4 (HPC-sensory cortex/SN/VAN), suggesting a relationship between vivid visual imagery and increased co-activation of this network. The persistence of CAPs was also found to be linked to TR-IM properties. Reliving was positively associated with the persistence of CAP2 (pHPC-visual cortex), suggesting that a more persistent activation of this network is associated with the experience of reliving the trauma. Emotional intensity showed a negative correlation with the persistence of CAP1, again highlighting the inverse relationship between emotional intensity and the resting-state pattern. Notably, no significant correlations were observed between CAP metrics and conventional clinical measures such as TR-IM frequency or retrospective recall, demonstrating the unique contribution of EMAs in elucidating the neural correlates of specific TR-IM properties. These findings suggest that distinct patterns of hippocampal-cortical co-activation are associated with various phenomenological properties of TR-IMs, offering new insights into their neurobiological underpinnings.

This study provides novel evidence for the distinct neural correlates of specific phenomenological properties of TR-IMs. The findings support the functional specialization of the anterior and posterior hippocampus in processing different aspects of these memories. The negative association between emotional intensity and the frequency/persistence of aHPC-DMN co-activation suggests a potential disruption in the normal regulatory mechanisms of affective memories, contributing to the spontaneous emergence of emotionally charged intrusions. The positive correlation between sensory features and the frequency of HPC-sensory cortex/SN/VAN co-activation supports the dual-representation theory and highlights the role of sensory processing in the vividness of these memories. The association between reliving and the persistence of pHPC-visual cortex co-activation suggests a potential mechanism for the “here-and-now” experience. The lack of association between CAPs and conventional symptom measures underscores the importance of utilizing ecologically valid assessments like EMAs to capture the nuances of the TR-IM experience. This study provides evidence for a complex interplay between different brain networks in the generation of TR-IMs. The results suggest that therapeutic interventions targeting these dynamic hippocampal-cortical network interactions might offer novel approaches for treating this debilitating symptom.

This study demonstrates the importance of considering the unique phenomenological properties of TR-IMs when investigating their neural substrates. The findings highlight the distinct contributions of aHPC-DMN and pHPC-visual cortex interactions to the emotional and sensory aspects of these memories. The study supports the use of ecological momentary assessments for capturing the complexity of TR-IMs and suggests potential therapeutic targets for interventions aimed at reducing the burden of these distressing memories. Future research should explore the use of task-based fMRI studies to investigate the dynamic neural activity during memory retrieval and explore the role of other sensory modalities beyond vision.

The study's limitations include the use of resting-state fMRI, which might not fully capture the neural dynamics during active memory processing. The cross-sectional design limits the ability to make definitive causal inferences. The predominantly female sample size might limit the generalizability of the findings to males. The ecological momentary assessments were completed over a two-week period, and it's possible some memory details could be lost over this timeframe. Future studies could address these limitations by using task-based fMRI, longitudinal designs, and balanced samples.