Dissociative amnesia (DA) is a psychiatric disease in which patients experience a failure to recall important autobiographical memories in the absence of any known neurological or other medical condition (1). Due to the imprecision of the psychiatric definition, DA encompasses a broad group of clinical presentations (2). Some patients experience a focal loss of memory encompassing a limited time-period (e.g., 5 years of their life) or a specific theme (such as a relationship or a work experience), most of the time stressful. However, some patients exhibit a more spectacular display of their symptoms with a global amnesia encompassing their whole lifespan and even sometimes a loss of identity and a fugue.

Dissociative amnesia often appears after mild traumatic brain injuries (3) or psychological stressors (4). Its prevalence rates are estimated between 0.4 and 7.3%, with a sex ratio of 1:1 (5). It occurs mostly between the age of 20 and 40 years old, even though cases among children and older people have been described (5). Its pathophysiology has not been elucidated yet. Nonetheless, two theories have arisen in the past decades. Kopelman’s theory of the “stress-related explanatory model” (6, 7) stated that patients suffering from DA were more likely to have prior biopsychosocial vulnerability, with a weaker cognitive executive reserve, leading to an inability to recall old memories. According to this theoretical model, by acting on executive functioning, stress may weaken the ability to retrieve their autobiographical memories (8). More recently, Markowitsch and his colleagues hypothesized that the memory recollection being blocked was under the influence of a fronto-temporal desynchronization due to a dysfunction in the hypothalamic-pituitary-adrenal (HPA) axis (9). The second pathophysiological lead (namely the “two-hit hypothesis”), which does not exclude the first theory, assumes that patients with DA suffer from an additive interaction between both physical and psychological factors causing the amnesic state (10). To date, its treatment relies on the management of the potential underlying depressive disorder, the identification and resolution of the eventual underlying crisis and a contextual interview of memories progressively moving from earlier and less stressful events to more recent ones (2). Contemporary psychotherapeutic techniques include cognitive behavioral therapy and acceptance/commitment therapy (11).

To our knowledge, no previous review has focused specifically on the functional brain alterations observed when using neuroimaging in patients with DA. In the present systematic review, we gathered and analyzed the data available to draw future perspectives in the field of neuroimaging studies and in the treatment of DA.

We conducted a systematic review of the international scientific literature using the bibliographic search engine PubMed. The following medical subject headings were used, accordingly to the possible denominations of DA developed in Staniloiu’s review article (5): (“hysterical amnesia” OR “dissociative amnesia” OR “dissociative fugue” OR “psychogenic amnesia” OR “functional amnesia” OR “mnestic block syndrome” OR “medically unexplained amnesia” OR “idiopathic amnesia” OR “disproportionate persistent retrograde amnesia” OR “fugue” OR “focal retrograde amnesia”). These were combined with keywords related to functional neuroimaging (“neuroimag*” OR “MRI” OR “fMRI” OR “PET” OR “SPECT” OR “imaging”).

Our inclusion criteria were the following: studies published up to December 1, 2022, involving patients with DA according to the Diagnostic and Statistical Manual of Mental Disorders (fifth edition or prior) who had undergone at least one brain imaging procedure (regardless of the technique used) while they were amnestic.

We searched the database using a predefined search algorithm to identify potentially eligible studies. Articles in which patients had isolated anterograde amnesia with no retrograde amnesia were withdrawn because of the probable different brain mechanisms implicated in their pathophysiology (12). We discarded studies involving EEG investigation because of the poor spatial resolution of this technique. Studies only involving structural techniques were also put aside as they did not add any functional information. All duplicate studies were removed. We independently selected studies based on their titles. Then, all online abstracts were reviewed, and full-text papers were retrieved when relevant. The procedure followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis criteria (PRISMA) (13).

We found 310 studies. 32 were duplicates. Among the 278 remaining, 51 were reviews, and 109 were considered irrelevant either because of their title (17) or their abstract (18). Through the 118 last, 94 lacked imaging analysis and one was about a pure anterograde amnesia. Also note that one study was discarded (19) because the subjects it described was presented elsewhere (18). Finally, we identified 22 studies (one large-sample retrospective controlled study and 21 case reports or case series) fulfilling our inclusion criteria, with a total number of 49 individual patients (see Figure 1, PRISMA flowchart of the review) (18, 20–40). Table 1 summarizes the main features (socio-demographic data and clinical features of the patients, experimental paradigms, delay to imagery) and results of the studies included in the review.

Numerous reports studied the neural mechanisms underlying the neural correlates of DA. Despite their heterogeneity, these reports predominantly showed a functional modification in the frontal areas, the limbic structures and the temporal lobe. Marginally, some reports showed abnormalities in the posterior and inferior parietal cortex, the occipital cortex and the cerebellum. Four of the six studies with another brain imaging technique after remission showed a normalization of the patterns observed when patients were amnesic. In the frontal regions of the brain, at rest, patients with DA mostly presented hypoactivity of the right inferolateral PFC, the right frontal gyrus and the bilateral frontobasal, frontomesial and orbitofrontal regions compared to HC (18, 27, 32). When performing a memory task, the patients with DA had greater activation in the frontal areas of their brains during recognition tasks, while their left dorsolateral PFC and right middle lateral frontal cortex failed to activate correctly during recollection. However, those results rely on uncontrolled studies and replication are need. Besides the frontal abnormalities described above, at rest, patients with DA showed diminished activation of their temporal areas compared to HC, especially in the left hemisphere (22, 28, 29, 32). Moreover, when they tried to recognize or recollect memories from their amnestic past, they failed to activate their bilateral hippocampi and para-hippocampi compared to HC (20), leading to lower BOLD signal in the studies reviewed. Finally, patients with DA had decreased activation of their limbic system (noticeably in the left insula, amygdala, and putamen) at rest compared to controls (28) but demonstrated broad abnormalities during memory tasks. They showed hyperactivity of their amygdalo-hippocampal complexes and their ACC during the recognition process compared to HC (20–22), while these regions were deactivated when the patients tried to recollect memories. The only common point between pure GARA and pure FRA patients was a decreased activity in the temporal lobe at rest. The comparison between acute and chronic states only highlighted a shared pattern of decreased rCBF through the temporal lobe (encompassing hippocampi and amygdalar structures). These data seem to favor an involvement of the temporal lobe as a core node of DA, independently of the duration of the symptoms or the time span encompassed by the amnestic state. Noticeably, three studies gathering eight patients reported normal SPECT examinations, but the results were controlled to HC in only one subject.

Three core brain networks that play a key role in the coordination of cognitive, affective and interpersonal processing have been identified: the default mode network (DMN), the central executive network (CEN), and the salience network (SN). Patients with DA show large anatomically dispersed abnormalities in each one of these three networks (see Figure 2). The DMN is constituted of functionally correlated brain areas (PCC, ventromedial PFC, medial, lateral, and inferior parietal cortex) (42). Among other functions, it is implicated in the interplay between emotional processing and cognition (43). Its increased function is demonstrated in self-related mental activity (44). Patients with DA had decreased activation of their ventromedial PFC (25, 27) and their posterior parietal cortex (32) at rest compared to HC. CEN, on the other hand, is a network antagonist to the DMN; this network is anchored in the bilateral dorsolateral PFC and the lateral posterior parietal cortex. It tends to activate in emotional and higher-order mental states (45). In the studies reviewed, the activation of the CEN was different depending on whether the patient was intensively recollecting memories or only recognizing them. In recognition, the BOLD signals observed in their dorsolateral PFC and their lateral posterior parietal cortex were higher than in HC (32, 35), while the dorsolateral PFC failed to activate during recollection (33). Lastly, the dynamic switch between the arousal of the DMN or the CEN is dependent upon the SN. More precisely, saliency detection has prominent consequences on how internally- (DMN) and externally-directed (CEN) cognitions are processed by one’s conscience (46). The SN is a large-scale network whose two major nodes are the dorsal part of the ACC and the anterior insulae, but it also includes the amygdala, the ventral striatum and the ventral tegmental area. Two patterns of SN activation–depending on the memory task performed (recognition or recollection)–emerge from the studies included. During recognition, patients with DA had a diminished rCBF in their insulae (36), and an increased BOLD signal and rCBF were observed in the ACC (22, 31) and the amygdala (20, 31) compared to controls, while both their insulae and ACC failed to activate during recollection (33, 39). Based on these observations, a network-centered approach may be useful to better understand the neural basis of DA and to gather the sparse results of the scientific reports previously reported.

Approximately three out of four DA are related to an identified traumatic event (4), establishing a strong link between stress and this disorder. Memory suppression is a coping strategy that helps individuals to suppress emotional memories through two time-differentiated procedures (47): the first implies the progressive deactivation of the right inferior frontal gyrus, the pulvinar and the fusiform gyrus, while the second is linked to the decreased activity of the amygdalo-hippocampal complex, which is anti-correlated to the right medial frontal gyrus’ activation over time. Patients with post-traumatic stress disorder (PTSD) have increased activation of their dorsolateral PFC (48) and increased functional connectivity between the inferior frontal gyrus and the para-hippocampus during experimental memory suppression tasks (49) compared to controls. Phenomenologically speaking, while PTSD could reflect the failure of the memory suppression system to remove unwanted recollections from the patient’s mind, DA might, on the contrary, be considered an involuntary overactivation of this system, eliminating the ability to retrieve any autobiographical episode (see Table 3). Several studies documented a dysfunction in brain regions involved in memory suppression in DA, such as the right dorsolateral PFC (35), the right inferior lateral frontal cortex (18, 36), the hippocampi and the para-hippocampal regions (20, 31). One study used a think/no-think paradigm to study the involvement of memory suppression in DA (30). However, because of its design (a single case experiment), it is difficult to draw any firm conclusion about the involvement of the memory suppression system in DA.

Dissociative states are still poorly understood. The two other main clinical entities alongside DA in the dissociative disorders chapter in the DSM are dissociative identity disorders and depersonalization/derealization disorders. Patients with dissociative identity disorder have smaller hippocampal (17, 50) and amygdalar (17) volumes than controls, an decreased rCBF in the bilateral orbitofrontal cortex and an increased rCBF in the median and superior frontal regions and the occipital areas bilaterally (51). Patients with depersonalization and derealization disorders have a lower fractional anisotropy than HC within the right temporo-parietal junction and the left temporal lobe (52). Their right ventrolateral PFC seems to be hyperactivated in depersonalization disorder, with a leading role of this structure in the “top-down” inhibition of emotional responses (53). The constellation of brain areas linked to dissociative experiences previously identified in the literature overlaps with the brain regions identified as dysfunctional in patients with DA. Noticeably, this includes brain areas involved in the DMN (bilateral orbitofrontal cortex, posterior parietal cortex) and the hippocampi, which are of interest in DA, as described above.

Yet, despite some converging results, several limitations appeared. First, the population included might not be representative of DA patients. Indeed, while we expected a 1:1 sex ratio based on the epidemiology, most of the subjects included in the studies were men. Although this may be linked to a probable sampling bias due to the limited number of patients included in the review, it is possible that sex plays a role in the results. Secondly, patients included in the review were very heterogeneous in term of clinical features (ranging from GARA to FRA and sometimes LOI) and comorbidities. It is obvious that the actual presence of an active comorbidity such as depression or functional neurological disorder may have influenced the results of the studies. We tried to compare GARA and FRA patients but failed to draw any conclusion because of the disparate imaging techniques employed to assess the brain modifications, and we were not able to analyze imaging data based on a LOI grouping because of the heterogeneity of such a thing. This heterogeneity adds another constraint to the generalization of these data. The same comment applies to the delay between the onset of symptoms and the imaging procedure. It is worth noting that over time, the changes observed in amnestic patients may be different because of mechanisms such as neural plasticity, re-learning their own biographical information or even social, familial or behavioral reinforcers. We also tried to compare the data on chronic and acute states, but once again, the comparison was made difficult by the numerous paradigms employed. Finally, it is possible that the sole fact of observing the phenomenon might influence the results, known as the Hawthorne effect. In most cases, controlling the results to those of HC might be helpful to limit this effect, but only one of the studies included in the review presented data compared to a control population. Our review emphasized the lack of prospective controlled studies in the field of neuroimaging in DA. The overall quality of the 22 studies included was poor and some of the most obvious confounding factors were not often considered such as history or active neurological or psychiatric condition, or medication for example (see Table 2). Additionally, the methods varied greatly, both in the neuroimaging technique employed and the tasks performed (see Table 1). It is thus exceedingly difficult to draw any firm conclusions. Surprisingly, we failed to find a single study using a resting-state fMRI paradigm. Resting-state fMRI is a promising tool to study brain function, and further studies employing this technique seem mandatory in the years to come to help better understand the pathophysiology of DA.

Progress made in the field of non-invasive brain stimulation techniques were very important over the past decades. It is now commonly used to treat some classic psychiatric conditions such depressive disorders (54) or refractory hallucinations (55). Moreover, recent experimental approaches also demonstrated their efficacy in neurological-like psychiatric disorders such as functional weakness (56) and functional motor disorders (57). Techniques like repetitive transcranial magnetic stimulation or transcranial direct current stimulation have also been used to enhance cognitive functioning, and a recent review revealed that non-invasive brain stimulation had a small but significant pre-cognitive effect on attention and working memory (58). A better description of the brain areas involved in DA may be of use to determine new targets for non-invasive brain stimulations and eventually be a future lead in the treatment of this neuropsychiatric condition.

Dissociative amnesia is an uncommon neuropsychiatric condition. Its pathophysiology is unclear, but our review in the field of neuroimaging studies suggests abnormal functioning of the DMN at rest and of the SN and the CEN when patients attempt to retrieve memories. The underlying mechanism might involve an overly activated memory suppression system, but further studies are needed to confirm this theory. The scientific literature about DA lacks methodologically strong studies, and despite numerous reports published in the field of neuroimaging in DA, only one of these publications is a prospective and controlled study, the remaining articles being, for the most, case reports. Moreover, most of the studies included used molecular imaging and task-performing fMRI. More recent approaches in the field of neuroimaging like resting-state fMRI could be of use to define more precisely the discrepancies among brain networks in future experiments. As very few forms of care management have scientifically shown their efficacy for DA yet, the unveiling of these in vivo mechanism using functional imaging may help scientists to define novel targets for neuromodulation treatments in this neuropsychiatric disorder.

ST, AY, BL, PP, and JP contributed to the conception of the study and made significant participation upon manuscript draft. ST, AY, and JP acquired the data (reviewing process). All authors contributed to the article and approved the submitted version.