Resilience to stress and trauma: a narrative review of neuroimaging research

In this narrative review, we summarise key findings from 38 human neuroimaging studies of resilience to stress and trauma, the majority of which have utilised structural and functional magnetic resonance imaging methods. Although prior research focused on overall differences between healthy controls and survivors of adversity, more recent studies have revealed neural markers unique to those who not only survived a potentially traumatic event but also maintained or regained their mental health following the event, thereby displaying a level of resilience. Such markers allow us to understand resilience at the level of the brain, to predict responses to adversity, and to measure outcomes of therapeutic interventions. The current body of evidence points to brain regions known to be affected by stress and trauma, including the prefrontal cortex (PFC; including the inferior and middle frontal gyri, dorsolateral and ventromedial PFC), anterior cingulate cortex, hippocampus, and amygdala, which implicates them as markers of resilience. Further research is needed to identify additional neural substrates of adversity and resilience, to confirm known markers, and to determine their full implications for mental health and recovery.


Introduction
In the context of stress and trauma, resilience is the phenomenon whereby individuals maintain or regain their mental health following adversity [1].Research on resilience aims to determine why some people recover from hardship more readily than others.Brain imaging research can reveal possible answers to this question by identifying neural markers of resilience [2].The purpose of this review is to summarise and discuss the resilience literature in terms of mental health and neuroimaging outcomes in humans.
In many studies of the neural correlates of adversity, only two groups of participants are compared.Some studies compare those who have experienced severe adversity with those who have not, while other studies compare those who were traumatised by severe adversity with those who experienced the same adversity but were not traumatised.Such studies have served as valuable springboards for subsequent resilience research, but as Zilcha-Mano et al. [3] have pointed out, at least three groups of participants must be compared in order to identify neural markers of resilience.This is partly because those who have experienced a potentially traumatic event (PTE) will exhibit brain alterations even if their mental health was not diminished by the event, and those with diminished mental health will not necessarily report any PTEs in their lifespan [4].Therefore, brain differences between PTE survivors and healthy controls (who have not been trauma exposed) do not necessarily reflect trauma or distress, and may reflect merely the occurrence of the PTE itself.Likewise, brain differences between traumatised and nontraumatised PTE survivors may reflect the difference in mental health status or the presence/absence of resilience, but it is unclear which.Ideally, comparisons across four groups of participants would reveal neural correlates distinctly related to resilience.These groupings would consist of ED (Exposed to adversity and Distressed), EH (Exposed to adversity but Healthy), UD (Unexposed to adversity but Distressed), and UH (Unexposed to adversity and Healthy) participants, as illustrated in Figure 1.With UH individuals (commonly described as 'healthy controls' in the literature) as the reference group, we might observe neural differences that differentially depend on exposure status and mental health status.This figure illustrates that two groups can share the same neural marker, so a third group (and ideally a fourth) is needed to identify unique neural markers for resilience.For example, neural marker A indicates PTE exposure but not trauma or distress, neural marker B indicates distress symptoms but not PTE exposure, and neural marker C indicates either a vulnerability factor or the lack of a resilience factor.(Note: other neural differences may exist too, e.g. to explain variations in distress between the two unexposed groups.)Thus, the EH and ED participants have marker A in common because both groups were exposed to PTEs (unlike the other two groups, which do not exhibit this marker), but they differ in terms of marker B (indicating distress) because the EH participants are not distressed.
Although there are four groups in this framework, when the mental health outcome is post-traumatic stress disorder (PTSD) symptoms, studies include only three groups (omitting the UD group) because PTSD symptoms stem from PTE exposure by definition.In such cases, neural markers B and C may not be distinguishable, and so neural marker D (unique to the EH group) may be examined as a possible resilience factor.Given these considerations, in the present review, we include studies of between-group comparisons involving at least three groups of participants (as defined in Figure 1).The review is divided into four sections based on the methods employed: cross-sectional group comparisons, cross-sectional linear models, longitudinal measurements, and intervention studies.Within each section, the relevant findings are organised by neuroimaging method.(Note: In reference to these methods, 'MRI' refers to structural magnetic resonance imaging, 'fMRI' refers to functional magnetic resonance imaging, and 'MEG' refers to magnetoencephalography.) Table 1, which summarises the 38 reviewed studies, is also structured in this way.

Group comparison studies (cross-sectional)
For brevity, we note that all the group comparison studies mentioned in this section employed a three-group design (i.e.UH, EH, and ED participants) unless stated otherwise and that all neural markers of resilience (i.e.markers unique to EH participants) were measured in comparison to the other groups under study (i.e. the UH and ED participants, as well as the UD participants in four-group studies).

Magnetic resonance imaging findings (grey matter measures)
We identified two studies in our literature search which employed a four-group design.In 1,870 European adolescents, Burt et al. [5] found larger grey matter volume (GMV) in EH participants in the right middle and superior frontal gyri.Similarly, in a sample of 804 German adults, Brosch et al. [6] observed larger GMV in the left middle frontal gyrus for EH participants.
Ten studies utilised a three-group design, five of which reported regions that may represent neural markers of resilience.In 39 US women with or without a history of intimate partner violence, Fennema-Notestine et al. [7] reported smaller GMVs in the medial temporal lobe, including the right parahippocampal gyrus, for EH participants.Using connectivity analyses in a study of 317 US military veterans (wherein the EH group had remitted PTSD rather than no history of it), Sun et al. [8] found greater centrality of the right subcallosal gyrus and smaller centrality of the left precentral gyrus in the EH group.Zilcha-Mano et al. [3] classified 129 US adults into one of two clusters based on GMV, cortical volume, and thickness measures, with one cluster containing significantly more EH participants.This cluster was characterised by larger volumes in several regions, especially the thalamus and rostral middle frontal gyrus.In two studies of a single sample of 131 US children and adolescents, Morey et al. [9] reported larger GMV in the left amygdala and right hippocampus, while Sun et al. [10] found larger centralities (within structural grey matter networks) of the transverse frontopolar gyri and sulci in EH participants.Interestingly, the initial study reporting on this sample did not identify any correlates specific to the EH youth [11].Illustration of possible neural differences among four groups of participants with differing exposure to PTEs and differing levels of mental health, with healthy controls (UH) as the reference group.This figure illustrates that two groups can share the same neural marker (e.g.marker A), so a third group (and ideally a fourth) is needed to identify unique neural markers for resilience.Neural marker A indicates PTE exposure but not trauma or distress; neural marker B indicates distress symptoms but not PTE exposure; neural marker C indicates a vulnerability factor (or a missing resilience factor); neural marker D indicates a resilience factor.Four other studies reported null results when examining neural markers of resilience in EH participants [12][13][14][15].Three of these studies hypothesised a role of hippocampal volume or chemistry in EH groups but found no between-group hippocampal differences that were unique to the EH group [13][14][15].Furthermore, the aforementioned study by Fennema-Notestine et al. [7] also found no differences in hippocampal volume among the three groups, contrary to their hypotheses.
Finally, a sample of 35 pairs of monozygotic US male twins has been analysed in multiple studies, with one twin in each pair being a combat-exposed Vietnam veteran (with or without PTSD) and the other a nonveteran without PTSD.The original study by Gilbertson et al. [16] found smaller hippocampal volumes in ED veterans compared with EH veterans; however, this difference was also found in their unexposed twin brothers, with the brothers of ED veterans having smaller hippocampi.Therefore, a smaller hippocampus may reflect a pre-existing vulnerability to PTEs rather than post-traumatic damage, or conversely, a larger hippocampus may serve as a resilience resource.Additional published studies have examined other potential neural markers in this sample (e.g.[17][18][19][20]); however, it is beyond the scope of our brief review to summarise all of them here, especially considering the possible lack of generalisability of results from a small sample of Vietnam veterans (see Ref. [21]).

Magnetic resonance imaging findings (diffusion tensor imaging)
Two studies have identified structural markers in EH participants.In 342 US adults, Ohashi et al. [22] found lower nodal efficiency, lower degree centrality, and lower closeness centrality in the right amygdala of EH participants.Furthermore, those with higher nodal efficiency in the left inferior frontal gyrus (pars triangularis) displayed a stronger positive relationship between childhood maltreatment and depression/anxiety symptoms, suggesting that decreased structural connectivity may confer protection against atypical activity propagating throughout the brain.However, in a study of 81 Dutch police officers, van der Werff et al. [23] found higher fractional anisotropy values for the left corticopontine tract in EH participants, indicating greater white matter integrity (i.e.increased connectivity) within this region (although this study found no GMV differences unique to the EH group).Clearly, the role of structural connectivity within the resilience process requires further examination.

Functional magnetic resonance imaging findings (restingstate)
Two studies examining resting-state functional data have reported patterns of decreased connectivity metrics or negative coupling that are specific to EH participants.In 38 US women, Cisler et al. [24] found that EH participants were characterised by lower centrality of the right ventrolateral prefrontal cortex (PFC) and fewer connections between this region and other nodes within an emotion regulation network, as well as lower levels of clustering and efficiency in the dorsal anterior cingulate cortex (dACC).This region was implicated again in a study of 33 Dutch adults, which showed greater negative connectivity between the left dACC and the lingual and fusiform gyri for EH participants [25].

Functional magnetic resonance imaging findings (taskbased)
In the two studies that examined task-based functional activation, a study of 28 US adolescents found greater activation in the bilateral orbitofrontal gyri and the left insula and putamen for EH participants, compared to ED and UH participants, when reading emotionally negative words [26].However, no task-related activity unique to the EH group was observed in a study of 84 German women in response to trauma-related words during an emotional Stroop task [27].

Magnetic resonance imaging findings (grey matter measures)
Moving away from group comparison designs, one study has utilised regression to examine structural (grey matter) covariation between regions that may be associated with mental well-being.In 242 Australian adults, Park et al. [28] found two structural networks associated with resilience (i.e.high well-being despite early life stress, e.g.child maltreatment), which encompassed frontal and temporoparietal regions, including the insula.

Functional magnetic resonance imaging findings (restingstate)
In a study 218 Chicagoan youth (aged 12-14 years), Miller et al. [29] found that neighbourhood violence was positively associated with poor cardiometabolic health (across six indices) only among youth with lower connectivity within the central executive network (CEN).Furthermore, lower CEN connectivity predicted poorer self-regulation of eating behaviour (in the context of ad libitum candy consumption, which may be considered a poor behavioural coping strategy).Connectivity within the default mode network had a similar protective effect, but only for one of the six indices, while connectivity within the anterior salience network did not have any moderating effects.

Functional magnetic resonance imaging findings (taskbased)
Six studies have investigated linear associations between task-related neural activation and measures of resilience.
In 55 male Israeli soldiers who had recently completed intensive combat training, Lin et al. [30] found that a bias towards threatening stimuli (i.e.faster reactions to angry vs neutral faces) predicted more severe PTSD symptoms and lower activity in the hippocampus, but only among those with high trait anxiety.Furthermore, decreased deactivation (i.e.relatively higher activation) of the dACC predicted greater hippocampal activation, which in turn predicted bias away from threats.
In a study of 820 US young adults, Corral-Frías et al [31] found a stronger relationship between childhood trauma and present-day anhedonia among those who exhibited lower functional activity in the left ventral striatum (VS) in response to positive feedback during a card-guessing task.In an earlier study of 170 young adults from the sample used by Corral-Frías et al. [31], Nikolova et al. [32] found stronger negative associations between recent life stress and positive affect in those with lower right VS activity during positive feedback.However, graphical inspection of these associations reveals that higher reward responsivity was associated with lower positive affect at lower severities of recent stress, suggesting that VS activity was not consistently protective of positive affect.Further research is clearly warranted.
In two samples of US females (168 children/adolescents, 146 women in mid-adulthood), Petrican et al. [33] found age-dependent functional correlates of resilience to recent adverse life events.In the youth sample, accelerated functional development of somatomotor and control-relevant networks and delayed functional development of the default mode network and salience/ventral attention systems were associated with delayed puberty and decreased resilience.In contrast, the opposite pattern (i.e.delayed ageing of somatomotor/control and accelerated ageing of default mode/salience systems) was associated with delayed menopause and increased resilience in the adult sample.These findings were interpreted in line with the theory that coping in youth is facilitated by earlier maturation, whereas adults in midlife benefit from delayed senescence, at least in the short-term.Indeed, the functional correlates of resilience in midlife were positively associated with a genetic risk of Alzheimer's disease, suggesting that resilience to recent adversity may imply longer-term detriments.
Two further studies used stress-inducing manipulations in healthy populations.In 70 Chinese university students, Hu et al. [34] reported a faster decline in taskelevated cortisol levels among those who exhibited greater activation in the left putamen during a monetary incentive delay task.This putamen activity was also positively correlated with the intrinsic connectivity between the right hippocampus and left inferior frontal gyrus, which in turn was negatively associated with cortisol awakening response.Finally, in 30 US young adults, Sinha et al. [35] found that increased responsivity of the ventromedial PFC (vmPFC) to highly aversive images was associated with higher scores on an active coping questionnaire, less-frequent interpersonal conflict, and lower levels of both emotional eating and alcohol consumption.

Magnetoencephalography findings (resting-state)
Two MEG studies have reported patterns of decreased connectivity that may confer resilience.In 38 PTE-exposed Italian outpatients, Brunetti et al. [36] found a negative association between trait resilience and connectivity (in the beta frequency band) between the medial PFC and several nodes in the default mode network, the salience network, and various sensory motor regions.In 199 US military veterans, James et al. [37] observed a negative correlation between lifetime PTE exposure and whole-brain synchronicity only in those without PTSD.They also found that increased trauma predicted reduced local connectivity, especially within the right superior temporal gyrus, only in the non-PTSD group.

Magnetic resonance imaging findings (grey matter measures)
Longitudinal studies provide a unique opportunity to examine directional relationships between brain structure/function and recovery from adversity (or a lack of recovery, i.e. vulnerability).Our search yielded one structural MRI study of 210 Dutch police recruits before and after emergency aid training, which involved exposure to PTEs [38].Those with smaller baseline volumes in the left dentate gyrus of the hippocampus were more likely to report increased PTSD symptoms and negative affect at follow-up, controlling for PTE exposure between assessments.

Functional magnetic resonance imaging findings (taskbased)
From six longitudinal studies that used experimental tasks to probe brain activity related to resilience, several key regions have been identified as potential markers, with mixed findings.For example, the anterior cingulate cortex (ACC) has been reported in two studies of youth resilience.In 20 US adolescents, Masten et al. [39] found that activity in the subgenual region of the ACC, dorsomedial PFC, and middle temporal gyrus during simulated peer rejection was unrelated to concurrent depression but predictive of increased depression 12-14 months later.However, in a study of 151 US children and adolescents, Rodman et al. [40] found that greater baseline activation in the right dACC (as well as the right superior frontal gyrus) during a cognitive reappraisal task predicted lower depression at follow-up (approximately 2 years later), albeit only in those with a history of maltreatment.Furthermore, lower baseline left amygdala activation predicted lower depression at follow-up in the maltreated participants.
Two other longitudinal studies have investigated amygdala activity in relation to resilience.Admon et al. [41] assessed 37 new recruits to the Israel Defense Forces before and after their deployment as military paramedics (all of whom experienced at least one medical-related PTE in this period).Baseline amygdala reactivity (but not hippocampal reactivity) to medical imagery predicted increased post-traumatic stress symptoms at follow-up and smaller increases over time in the functional coupling between the hippocampus and vmPFC.Furthermore, in a subsample of 24 of the paramedics, Admon et al. [42] found that greater baseline amygdala activation in response to potential punishment (in a dominoes game) predicted greater PTSD symptoms at follow-up.
Further longitudinal studies have implicated frontal regions in the resilience context.In a subsample of 160 Dutch police recruits from the aforementioned study by Koch et al. [38], Kaldewaij et al. [43] found that lower baseline activity in the left anterior PFC in response to task-induced emotional conflict was associated with larger increases in PTSD symptoms over time.Indeed, trauma load (during police training) predicted increased symptoms only in those with lower prefrontal activation at baseline.Finally, in a sample of 70 Canadian adults who had recently experienced a PTE and had to mentally visualise it in the scanner, Daniels et al. [44] found that greater activation of right middle frontal gyrus, right inferior frontal gyrus, and right thalamus was associated with higher trait resilience.In turn, trait resilience was negatively associated with PTSD symptoms at 5-6 weeks and 3 months post-PTE, even after controlling for childhood trauma.

Functional magnetic resonance imaging findings (taskbased)
Two clinical studies have examined neural markers of resilience in the context of psychotherapy for posttraumatic stress.Fonzo et al. [45] randomly allocated 66 PTE-exposed US adults to either immediate treatment or a waitlist control group.Before this, the participants responded to various emotional stimuli during scanning.
Greater activation in the dorsolateral PFC, vmPFC, VS, and left anterior insula, as well as lower activation in the amygdala, predicted symptom reduction in the treatment group but symptom elevation in the control group.Furthermore, greater activation in the dACC predicted symptom reduction in the treatment group and no significant symptom change in the control group.A subset of the treatment group was also randomised to receive transcranial magnetic stimulation of the right posterior middle frontal gyrus before treatment.In response, those who exhibited greater reductions in left amygdala activation reported greater symptom reductions post-treatment.Similarly, in 24 Brazilian police officers with partial PTSD (pPTSD), Peres et al. [46] observed decreased activation of the medial PFC and increased activation of the left amygdala in response to a real-life auditory recording of police officers dying under fire.Importantly, these changes in activation became less evident in the 12 pPTSD participants who received treatment, whereas the remaining 12 waitlist control participants did not show any change in neural response, and these neural changes mirrored the decrease in PTSD symptoms in the treatment group.Together, these clinical findings again highlight the important roles of the amygdala and various frontal regions, with reduced amygdala activity and increased frontal activity typically corresponding to greater resilience.

Discussion
In this brief narrative review, we have highlighted key neuroimaging studies that reveal plausible neural markers of resilience.Although this research pertains to a wide range of brain areas and connections, it yields recurring findings that accord with our current understanding of adversity and resilience.Among the most common neural signatures of maltreatment or trauma are alterations in regions and circuits that underpin memory (hippocampus), emotion processing (amygdala), reward sensitivity (VS), and executive function (PFC, dACC; see Refs.[4,47]).Therefore, it is unsurprising that resilience to adversity is related to the same neural regions, connections, or circuits.Regarding grey matter structures, our review indicates that resilience may be characterised by larger volumes in the hippocampus [9,16,38] and various frontal regions, such as the middle/superior frontal gyri [3,5,6].Regarding functional activity, resilience may involve greater activation of frontal regions such as the PFC [26,35,43,45,46] and inferior/middle/ superior frontal gyri [40,44], greater activation of the dACC [30,40,45], greater activation of the VS [31,32,45], and lower activation of the amygdala [40][41][42]45,46].These findings cohere with our theoretical understanding of the relevant brain regions.For example, larger and more active frontal areas may facilitate topdown regulation of negative emotions (e.g. via cognitive reappraisal), larger hippocampi may allow better processing of aversive or post-traumatic memories, greater activity in the VS may facilitate motivation-related adaptive behaviours despite adversity, and reduced activity in the amygdala may correspond to lower levels of or attenuated reactions to negative emotion or threat perception.The research is less definitive with regard to connections among such regions, but there is evidence that lower connectivity may underpin resilience in some contexts (e.g.[22,24,36,37]).As mentioned earlier, reduced connectivity might prevent dysfunctional activity from propagating widely in the brain [22].Alternatively, lower connectivity might reflect adaptation to adversity, whereby segregated networks allow flexible and more efficient responses that do not require remodelling of the entire functional system (see Ref. [47]).However, there is evidence that stronger connectivity within other networks or structures (e.g. the CEN, corticopontine tract, and connections between the hippocampus and frontal areas) may also characterise resilience [23,34,41].
Further research is clearly needed; indeed, even if we had identified every neural marker of resilience, there would still be the question of what these markers actually indicate or reflect.For example, if a given marker corresponded solely to an innate disposition (e.g. a genetically determined protective trait), then it might be relevant for prognosis or biologically-targeted intervention.Conversely, if the marker indicated an acquired ability (e.g.cognitive reappraisal techniques learned via psychotherapy), then it might inform efforts to develop treatments or other behavioural interventions (e.g. it could serve as an objectively measurable outcome).In addition, there are other gaps in the literature that can be addressed by future scientists.Most of the existing research is cross-sectional in design, and most of the cross-sectional studies involve group comparisons rather than linear models.For some neuroimaging analyses, the researcher usually specifies discrete groups of participants in order to locate regional differences (either structural or functional) across the groups, but once these regions have been identified, the researcher is free to extract the relevant metrics for each participant and test a linear model with other measures of interest, without imposing categorisations on the data.As Herzberg and Gunnar [47] have pointed out, linear models provide more precision and prevent linear trends from being obscured through coarse-grained group comparisons.Beyond cross-sectional research, our review also highlights the relative lack of longitudinal or intervention studies.Such designs are sometimes impractical in the domain of stress and trauma (e.g. an experimenter cannot allocate participants to be traumatised, and longitudinal studies require foreknowledge of impending PTEs), but they are not impossible, and the additional causal clarity that some such designs provide makes them especially valuable.On this point, it is worth mentioning the importance of studies in animals.While our brief review is limited to research on humans, animal studies constitute vital sources of insight on resilience, especially in experimental contexts.Such research should be consulted alongside the studies of humans summarised in this review.
A further gap in the literature pertains to the types of adversity and mental health outcome that have been covered to date.Understandably, researchers have predominantly focused on the most severe forms of adversity (e.g.childhood maltreatment, military service) and the most serious outcomes (e.g.PTSD), but the field of resilience research is much wider in scope.Of the studies we reviewed, only a handful included mental health outcomes from the positive domain, such as mental wellbeing [28], positive affect [32], active coping [35], or personal strengths [5].Even among the studies that focused on negative outcomes, relatively few considered outcomes other than PTSD symptoms.As mentioned earlier, PTSD studies cannot employ four-group designs, unlike studies of other forms of psychopathology (e.g.depression, anxiety).Beyond this methodological issue, it is worth investigating the full range of negative outcomes that may follow from stress or adversity, including behavioural ones such as problem drinking [26] or emotional eating [29].Likewise, future research should consider less extreme forms of adversity (e.g.stressful life events) in order to develop a fuller understanding of the neural underpinnings of resilience.Finally, even when a neural marker of resilience has been discovered, there remains the crucial question of whether the marker represents a persistent psychological advantage (e.g. a positive coping style) as opposed to a temporarily beneficial adaptation that may bring longerterm costs (e.g. the brain networks discussed by Petrican et al. [33]).Further research is needed to reveal the full time-course of the resilience process at the level of the brain.
There is also considerable scope for future neuroimaging research on the physiological mechanisms that underpin the psychological phenomenon of resilience and possibly result in the neural differences between resilient individuals and others.In this review, we have highlighted such differences that may explain why certain individuals recover more readily from adversity in terms of their individual mental health.But it would also be valuable to explore whether these neural differences relate not only to self-reported mental outcomes but also to underlying physiological processes that manifest as homeostasis and allostasis (see Refs. [48][49][50]).For example, it may be that a particular neural marker is correlated with both behavioural (e.g.adaptive coping strategies) and physiological (e.g.respiratory sinus arrhythmia [51]) indicators of resilience and adaptation.
In conclusion, this review has showcased key studies of the neural markers that uniquely characterise resilience to stress and trauma and has highlighted gaps in the literature that warrant further attention.With ongoing advances in neuroimaging techniques and scientific methods, we anticipate not only growth in our knowledge but also new interventions to promote resilience and recovery in those who have experienced adversity.

Figure 1 Current
Figure 1

studies MRI findings (grey matter measures) Author
Type of adversity/PTE Mental health outcome Resilience-related brain regions Koch et al., 2021 Police training PTSD symptoms, negative affect Hippocampus (left dentate gyrus)

Table 1
(continued ) a Regions empirically found to distinguish resilient participants from others.A table entry of 'none' indicates that the study did not find any markers unique to resilience.