Bipolar disorder is a chronic and severe mental illness with worldwide effects. It includes both the more severe bipolar disorder type I (BD I, requiring a manic or mixed episode) and the less severe bipolar disorder type II (BD II, requiring presence of a hypomanic episode).¹² While BD I affects men and women in equal numbers, BD II, like major depressive disorder, affects women in greater numbers. Estimates of lifetime prevalence of bipolar disorder in the general population range from 2.8% to 6.5%.¹³,¹⁴

Understanding and clarifying the diagnosis of bipolar disorder requires comparison to major depressive disorder since both include major depressive episodes.¹⁵ Previous research has focused on categorical differences between the two without considering that depressive episodes in major depressive disorder may or may not be the same clinically as depressive episodes that occur within the context of bipolar disorder.¹⁶ Additionally, differences may exist in the functional brain systems that underlie these disorders. Thus, it is important to examine both trait and state.

In medicine, trait generally refers to the formal diagnosis (major depressive disorder or bipolar disorder), which is relatively stable. State is the status of the patient at a given point in time (e.g., manic, depressed, euthymic). Thus, it is fluid and changeable. Recent advances in neuroimaging that utilize functional imaging techniques can contribute to our knowledge base of how manic and depressive states differ. This may help develop a better understanding of etiological and treatment differences between the two disorders.

Neuroimaging studies of bipolar disorder have found conflicting results in regard to structural abnormalities. General areas of focus include limbic structures and the prefrontal cortical pathways associated with the limbic system.¹⁷ The decision to focus on these areas seems, in part, motivated by research documenting abnormalities in the amygdala and prefrontal areas in major depressive disorder.¹⁸,¹⁹ In general, structural studies have found abnormalities in amygdala size in bipolar disorder patients.²⁰ However, some studies have found amygdala enlargement and others have found reduced amygdala volume.²¹,²² The sex of research subjects may also be a complicating factor, as activation studies have found male/female differences in anterior cingulate and dorsolateral prefrontal cortices as well as in the amygdala.²³ Since the proportion of men to women varies considerably across studies, this confounding factor may contribute to contradictory findings.

Some studies have focused on abnormalities in gray and white matter. A recent magnetic resonance imaging (MRI) study found abnormal gray matter density in the cingulate cortex and fronto-limbic cortex of patients with bipolar disorder who had poor outcomes on traditional treatment measures.²⁴ Reduced gray matter density in prefrontal regions and increased gray matter in temporal regions, including the amygdala, have also been documented.²⁵ Magnetization transfer imaging and voxel-based morphometry have shown small white matter density abnormalities in the anterior cingulate and subgyrus of bipolar disorder patients.²⁶ Diffusion tensor imaging has shown abnormal frontal white matter tracts in the prefrontal regions of bipolar disorder patients.²⁷ Gray matter volume was reduced in unmedicated bipolar disorder subjects in the posterior cingulate/retrosplenial cortex in a voxel-based morphometry MRI study.²⁸ Though specific findings are inconsistent, limbic and prefrontal regions are of interest in terms of understanding the neuropathology of mood disorders.

It is unclear whether the structural differences in the limbic and prefrontal areas in bipolar disorder have been highlighted because of unique differences in those areas or as a secondary function of the limbic system’s theoretical role in emotional processing. Theories regarding the limbic system (e.g., amygdala, thalamus, hippocampus) and emotional processing predate imaging advances in neuroscience. While perception of and reaction to fear has been clearly linked to the amygdala, other emotions have shown activation in various aspects of the limbic system as currently defined.²⁹ Discussion of the controversial definition of and role of the limbic system is beyond the scope of this article.

Several factors need to be considered prior to examining bipolar disorder imaging findings. Comparison of imaging studies of bipolar disorder is complicated by the different methodological techniques employed and by individual differences within the bipolar disorder population. Imaging studies of any cohort can involve some loss of accuracy when data are averaged across the groups. There is an additional complicating factor of patient state in mood disorders at the time of the scan. It is difficult to capture and maintain a manic state during functional imaging. There is both individual variance and variance between studies in how patient state is defined and measured. Brain function during a major depressive episode may differ from brain function when not depressed (i.e., euthymic), and both may be different from brain function in a healthy individual. In addition, patients with bipolar disorder who are in a euthymic state may be experiencing manic or depressive symptoms at a subclinical level.³⁰,³¹ Functional imaging studies controlling for these state differences in bipolar disorder will serve an important role in clarifying diagnosis and pathophysiology of bipolar disorder compared to major depressive disorder.⁷,³²

Medication differences within and between participant subgroups are another source of complexity for functional imaging studies. For example, the introduction of an 8-week trial of antidepressant medication has been shown to significantly reduce limbic activation in major depressive disorder patients compared to healthy subjects.³³ Many studies, especially of bipolar disorder populations, include patients on a variety of psychotropic medications. This limits comparability both within and between groups. A few studies have avoided this complication by focusing on unmedicated populations. This limits applicability to a clinical setting where the majority of bipolar disorder and severe major depressive disorder patients would likely be medicated. Medication factors should be considered when reviewing results across studies.

With the above caveats in mind, a few recent positron emission tomography (PET) studies have provided state and trait comparisons in bipolar disorder patients (Figure 1). One study compared regional cerebral blood flow (rCBF) between medicated type I bipolar disorder patients in euthymic and manic states.¹ The only significant differences were higher rCBF on the left in dorsal anterior cingulate cortex and caudate (head) in mania. Two studies (one measuring rCBF, the other measuring regional cerebral metabolic rate, rCMR) have compared bipolar disorder patients in euthymic and depressed states.⁴,⁵ In both studies, the patient groups were not significantly different. Reduced activity (greater reductions in the depressed state) has been noted in multiple cortical areas (dorsolateral prefrontal, medial prefrontal, orbital prefrontal, parietal) when bipolar disorder patients in all three states were compared to healthy individuals.²,⁴,⁵ Studies have reported increased activity in cerebellum in bipolar disorder patients compared to healthy individuals.⁴,⁵ One study also reported increased subcortical activity (ventral striatum, thalamus, right amygdala) in the depressed state.⁵ Subgenual cortex rCBF has also been noted to be decreased in depression (both in bipolar disorder and major depressive disorder) and increased in manic states compared to normal.³ Euthymic bipolar disorder patients had increased rCBF in dorsal anterior cingulate cortex when compared to healthy individuals in one study.⁴ Increased rCMR in the left amygdala has also been reported in bipolar disorder patients in the depressed state, and in the unmedicated euthymic state.³⁴ Interestingly, this was not the case for euthymic bipolar disorder patients on mood stabilizing medications. From these studies, it appears that the manic state is associated with increased activity in anterior cingulate regions while the depressed state is associated with decreased activity in medial, lateral, and orbital prefrontal regions and increased activity in associated subcortical structures. Studies of the euthymic state suggest a combination of the increased activity in anterior cingulate regions and decreased activity in other prefrontal areas, consistent with the presence of mild symptoms of both mania and depression.

A few studies have compared the depressed and euthymic states in bipolar disorder with major depressive disorder in an effort to detect trait differences. In one study, increased left amygdala metabolism was found in the depressed state for both conditions.³⁴ Another study used principal component analysis to group symptoms on the Beck Depression Inventory (BDI) into four clusters (negative cognitions, psychomotor-anhedonia, vegetative, somatic).³⁵ Intercorrelations between these components and correlations between BDI total and component scores with both absolute and normalized rCMR were evaluated for both major depressive disorder and bipolar disorder patients. The components were not intercorrelated for the major depressive disorder group, suggesting that they represent separate aspects of depression. In the bipolar disorder group, the negative cognitions, psychomotor-anhedonia and vegetative components were highly intercorrelated, suggesting a much more unified condition. Consistent with these findings, the brain areas in which both absolute and normalized rCMR correlated with each component were distinctly different in the major depressive disorder group, supporting the involvement of different neuronal networks for each symptom cluster. In contrast, for the bipolar disorder group there were no significant correlations for the negative cognitions and vegetative symptoms components using absolute rCMR. There was considerable overlap in brain areas for all components using normalized rCMR. The psychomotor-anhedonia component score had the strongest correlations with rCMR for the bipolar disorder group (stronger than any other component or the total BDI). The psychomotor-anhedonia component scores correlated with lower absolute metabolism in the insula, anteroventral basal ganglia and temporal and inferior parietal cortices. There was higher normalized metabolism in the anterior cingulate cortex. These results suggest considerably more variability of functioning within the subgroup of individuals diagnosed with major depressive disorder than bipolar disorder, and offer further support that depressive states in bipolar disorder are likely to be neurologically different than depressive states in major depressive disorder though both share clinical similarities.³⁵

Facial expression recognition tasks have been used in conjunction with functional imaging to probe prefrontal-limbic system-related performance. Individuals with bipolar disorder have been shown to have impaired ability to discriminate the intensity and valence of emotion. The identified deficits in emotion recognition are state-specific. Subjects in the manic state had significantly impaired recognition of fear and disgust in a task requiring recognition of the six basic emotions (fear, disgust, anger, sadness, surprise, happiness). The most common errors were mislabeling of fear as surprise and disgust as anger.³⁶ In addition, an inverse correlation was found between the intensity of manic symptoms and recognition of sadness.³⁶ Other studies have also noted impaired ability to recognize and estimate the intensity of sadness in the manic state.⁸,⁹

Functional MRI (fMRI) studies generally have utilized comparison of two emotional states (happiness, sadness or positive, negative). Individuals with bipolar disorder in the manic state have relatively normal patterns of activations to positive emotional states, but not in response to negative emotional states. Negative emotions evoke less cortical and more subcortical activation when the manic state is compared to normal, although the specific areas vary across studies (Figure 2).⁶—⁹,³⁷ Overall, these findings are consistent with the impaired ability of bipolar disorder individuals in the manic state to identify and estimate intensity of negative emotions. Future functional imaging studies designed to probe a wider range of emotional recognition, as is done in neuropsychological testing, will be needed to clarify the specific areas associated with particular impairments.

Impaired facial affect recognition has been demonstrated in euthymic individuals with bipolar disorder, although data on specific emotions were not presented.³⁸ Some studies focusing on the euthymic state fit nicely with those examining mania. In one, an inverse correlation was found between the severity of subclinical manic symptoms and recognition of anger and fear.³⁹ However, this study also found that euthymic bipolar disorder patients had significantly better than normal identification of disgust (more detected and fewer mislabeled). Another study found that euthymic BD II subjects had significantly better than normal recognition of fear, while euthymic bipolar disorder I subjects were slightly impaired.³⁶ In contrast, the BD II subjects were slightly impaired on recognition of disgust, but the BD I subjects were not. Brain activation correlates of these differences are not yet available, as functional imaging studies have reported results on the manic and depressed states only.

Neuropsychological studies of emotion recognition in depression have concentrated on major depressive disorder. As discussed previously, the assumption that major depressive disorder and bipolar disorder depressive states are identical is not supported. Little information is available about emotion recognition by bipolar disorder individuals in the depressed state. One recent study found bipolar disorder subjects in the depressed state no different from normal on recognition of sad, fearful, or happy faces, but more research is clearly needed in this area.⁹

Functional MRI studies have utilized comparison of two or three emotional states (happiness, sadness, fear or positive, negative). Individuals with bipolar disorder in the depressed state had more cortical and subcortical activations to positive emotional states than were seen in either healthy individuals or individuals with major depressive disorder.⁷,⁴⁰ Specifics varied across studies, but included areas in ventral and medial prefrontal and anterior temporal cortices (Figure 2). Negative emotions evoked less cortical and more subcortical activation than normal (Figure 2). This is similar to the response seen in the manic state, and so may be trait- (rather than state-) associated.

All of the above studies have focused on tasks that involve cognitive processing of experimentally selected emotional stimuli. One group has compared rCBF changes (measured by PET) in patients with bipolar disorder following induction of sadness. Sadness was induced by exposure to a personal script of negative life events.⁴,¹⁰ Bipolar disorder patients (both depressed and euthymic) differed from healthy individuals in having decreased rCBF in medial and orbitofrontal cortex and in not having increased rCBF in subgenual cortex (Figure 3). All three groups had increased rCBF in cerebellum and insula and decreased rCBF in parietal cortex. Healthy and depressed bipolar disorder groups both had decreased rCBF in lateral prefrontal cortex. Only the euthymic bipolar disorder group had increased rCBF in dorsal anterior cingulate and premotor cortices.⁴ Patterns of rCBF change were also compared between a euthymic bipolar disorder group and their healthy siblings.¹⁰ Overall the patterns of response were similar, with the exception of the medial frontal cortex. The healthy sibling group had increased (rather than decreased) rCBF in this area. The authors of the study suggested that the differences from the normal pattern in orbitofrontal and dorsal anterior cingulate corticies may be trait-related, as they were present in both the bipolar disorder groups and healthy siblings. Two euthymic bipolar disorder groups responsive to different mood stabilizing medications were also compared. The lithium-responsive group had fewer areas of decreased rCBF in medial frontal cortex, and increased (rather than decreased) rCBF in rostral anterior cingulate cortex (Figure 3). The authors noted that decreased activity in this region may indicate more severe illness, as it is believed to be involved in detecting shifts in affect and correcting emotional responses.

Functional neuroimaging of bipolar disorder is still at a relatively early stage of development, compared to neuroimaging data on other mental disorders. Multiple complicating factors have likely contributed to the relative paucity of research on bipolar disorder, including diagnostic difficulties related to subtypes and differentiation between the mood disorders. Additionally, capturing and measuring state immediately prior to scan and controlling for the medication variability further complicate neuroimaging studies. Affective tasks selected to study bipolar disorder tend to generally include a cognitive processing component. Further research in this area is needed, with specific focus on emotion induction, rather than emotion recognition tasks. Studies should consider, control, and report, as much as possible, data on vital patient variables, such as state at time of scan, medication, and comorbidities.